NBA Positions.Rmd

---
title: "NBA Positions"
author: "Olusegun"
date: "2021-04-24"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r}
rm(list=ls())

### Initialize data sets to work with
# Webscrape Statistics from Baskballreference
#library(rvest)
#thelink <- xml2::read_html("https://www.basketball-reference.com/leagues/NBA_2021_per_game.html")
#nba <- as.data.frame(thelink %>%html_nodes("table")%>%.[[1]] %>%html_table())

# Remove repeated header row
#unn <- subset(nba, nba$Rk != 'Rk')

# Remove duplicated player records (duplicated if player traded, keep only 'Total' record)
#nba2 <- unn[!duplicated(unn$Player),]

# Convert regressor variables to integer values
#i <- c(6:30)
#nba2[ , i] <- apply(nba2[ , i], 2, function(x) as.numeric(as.character(x)))

# Filter out players not having played >= [pct] of total games played
#pct <- .7 #define percentage of games player must have played in as compared to max player
#nba3 <- subset(nba2,nba2$G >= max(nba2$G)*pct)

# Remove attempts and attempts made columns (as they are simple divisors and dividends of percentage columns)
#nba4 <- nba3[-c(1,2,4:8,9,10,12,13,15,16,19,20)]
#summary(nba4)

################### CSV SAVED ON APRIL 8th 2021 ####################
#write.csv(nba3, "nba3.csv")

nba3 <- read.csv('/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/6440 project/Dataset/nba3.csv')
colnames(nba3)[11] <- 'FG%'
colnames(nba3)[12] <- '3P'
colnames(nba3)[13] <- '3PA'
colnames(nba3)[14] <- '3P%'
colnames(nba3)[15] <- '2P'
colnames(nba3)[16] <- '2PA'
colnames(nba3)[17] <- '2P%'
colnames(nba3)[18] <- 'eF%'
colnames(nba3)[21] <- 'FT%'

nba4 <- nba3[-c(1,2,4:8,9,10,12,13,15,16,19,20)]
summary(nba4)
```

```{r}
#nba4 <- na.omit(nba3) #remove records with 'na's
nba3$`3P%`[is.na(nba3$`3P%`)]<-0 
nba4$`3P%`[is.na(nba4$`3P%`)]<-0 

# Construct position specific dataframes and assign values to specific positions
table(nba4$Pos)
```

```{r}
nba3[nba3$Pos =='C-PF',] # Kelly Olynyk
```

```{r}
nba3[nba3$Pos =='PG-SG',] # James Harden
```

```{r}
nba3[nba3$Pos =='SF-SG',] # Sviatoslav Mykhailiuk
```

```{r}
nba3[nba3$Pos =='SG-PG',] # Delon Wright
```

```{r}
nba3[nba3$Pos =='SG-SF',] # Hamidou Diallo & Norman Powell
```

```{r}
################# KNN TO DEFINE PLAYERS ################# 
#### Olynyk: C OR PF?? ####
source("/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/R_codes/myfunctions.R")

knn <- subset(nba4,nba4$Pos =='C' | nba4$Pos =='PF')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
library(FNN)
library(caret)
```

```{r}
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "C")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "#FC4E07")
```

```{r}
max.k <- which.max(kappa)
max.k #1 is too low, use 13, next highest, instead
```

```{r}
max.k <- 13

### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="C-PF"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #Label as PF
```

```{r}
#### HARDIN: PG OR SG?? ####
knn <- subset(nba4,nba4$Pos =='PG' | nba4$Pos =='SG')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "PG")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "#FC4E07")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="PG-SG"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label as PG

```

```{r}
#### Mykhailiuk : SF OR SG?? ####
knn <- subset(nba4,nba4$Pos =='SF' | nba4$Pos =='SG')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "SF")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "#FC4E07")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="SF-SG"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label as SG
```

```{r}
#### Wright: SG OR PG?? ####
knn <- subset(nba4,nba4$Pos =='SG' | nba4$Pos =='PG')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "SG")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "#FC4E07")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="SG-PG"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label as SG
```

```{r}
#### Diallo & Powell: SG OR SF?? ####
knn <- subset(nba4,nba4$Pos =='SG' | nba4$Pos =='SF')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "SG")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "#FC4E07")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="SG-SF"),]
x0 <- x[1,][-1] #Diallo
x1 <- x[2,][-1] #Powell

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 
x1.scaled <- scale(x1, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn0 <- knn(train = training.scaled, test = x0.scaled,
            cl = training.data[,1], k = max.k)
Knn0
```

```{r}
Knn1 <- knn(train = training.scaled, test = x1.scaled,
            cl = training.data[,1], k = max.k)
Knn1
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn0)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) 
```

```{r}
table(knn$Pos) #label Diallo as SG MAJORITY RULE
```

```{r}
Knn.attr <- attributes(Knn1)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label Powell as SG
```

```{r}
#################### POP OUT OF KNN CLASSIFICIATION ###################
#define Olynyk as KNN result, viz. "PF"
nba4$Pos[nba4$Pos=="C-PF"] <- "PF"
nba3$Pos[nba3$Pos=="C-PF"] <- "PF"

#define Hardin as KNN result, viz. "PG"
nba4$Pos[nba4$Pos=="PG-SG"] <- "PG"
nba3$Pos[nba3$Pos=="PG-SG"] <- "PG"

#define Mykhailiuk  as KNN result, viz. "SG"
nba4$Pos[nba4$Pos=="SF-SG"] <- "SG"
nba3$Pos[nba3$Pos=="SF-SG"] <- "SG"

#define Write as KNN result, viz. "SG"
nba4$Pos[nba4$Pos=="SG-PG"] <- "SG"
nba3$Pos[nba3$Pos=="SG-PG"] <- "SG"

#define Powell as KNN result, viz. "SG"
nba4$Pos[nba4$Pos=="SG-SF"] <- "SG"
nba3$Pos[nba3$Pos=="SG-SF"] <- "SG"

Dat <- nba4
Dat2 <- nba3[-c(1,4:8,9,10,12,13,15,16,19,20)]

Dat$Pos <- as.factor(Dat$Pos)

table(nba4$Pos) #good
```

```{r}
################### CSV SAVED ON APRIL 11th 2021 ####################
#write.csv(Dat2, "NBAper.csv")

## Create training and test data ##
#source("C:/Users/super/Documents/MSA/Sp21/MSA 6440/Data/myfunctions.R")

#source("C:/Users/super/Documents/MSA/Sp21/MSA 6440/Data/myfunctions.R")
source("/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/R_codes/myfunctions.R")
RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0) ## set seed so that you get same partition each time
p2 <- partition.2(Dat, 0.7) ## creating 70:30 partition
training.data <- p2$data.train
test.data <- p2$data.test

nba.nolbl <- training.data[,-1]
nba.scaled <- scale(nba.nolbl, center = TRUE, scale = TRUE)

# https://predictivehacks.com/how-to-determine-the-number-of-clusters-of-k-means-in-r/
library(factoextra)
```

```{r}
library(NbClust)
library(mclust)
```

```{r}
set.seed(123)

fviz_nbclust(nba.scaled, kmeans, method = "wss") +
  geom_vline(xintercept = 2, linetype = 2)+
  labs(subtitle = "Elbow method")
```

```{r}
fviz_nbclust(nba.scaled, kmeans, method = "silhouette")+
  labs(subtitle = "Silhouette method")
```

```{r}
fviz_nbclust(nba.scaled, kmeans, nstart = 25,  method = "gap_stat", nboot = 500)+
  labs(subtitle = "Gap statistic method") # none of these charts give us good idea for number of Clusters
```

```{r}
############# Clustering ##################
#### Mclust #####
library(mclust)
RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)

mc.list <- list()
bic <- NULL
for (k in 1:10){
  mc <- Mclust(Dat[-1], k)
  bic <- c(bic, mc$bic)
  mc.list[[k]] <- c(mc.list, mc)
}

plot(seq(1,10,1), bic, type = "l", xlab = "Number of clusters")
```

```{r}
plot(bic[1:10], type = "l", xlab = "Number of clusters")
```

```{r}
maxbic <- which.max(bic[1:10])
maxbic
```

```{r}
mc3 <- Mclust(Dat[-1], maxbic)

table(mc3$classification, Dat$Pos)
```

```{r}
# Mclust Clustering Plots
library(tidyr)
library(ggplot2)

dat.mc <- as.data.frame(t(mc3$parameters$mean))
dat.mc$clusters <- row.names(dat.mc)

# Percentages Clusters
dat.mc1 <- dat.mc[c(1:5,15)]

dat.g <- gather(dat.mc1, stats, mc3mean, -clusters)

mc.1 <- ggplot(dat.g, aes(stats, mc3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Percentages") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.1
```

```{r}
# Offense Clusters
dat.mc1 <- dat.mc[c(6,9,14,15)]

dat.g <- gather(dat.mc1, stats, mc3mean, -clusters)

mc.2 <- ggplot(dat.g, aes(stats, mc3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Offense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.2
```

```{r}
# Defense Clusters
dat.mc1 <- dat.mc[c(11,10,7,15)]

dat.g <- gather(dat.mc1, stats, mc3mean, -clusters)

mc.3 <- ggplot(dat.g, aes(stats, mc3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Defense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.3
```

```{r}
# Special Clusters
dat.mc1 <- dat.mc[c(12,13,15)]

dat.g <- gather(dat.mc1, stats, mc3mean, -clusters)

mc.4 <- ggplot(dat.g, aes(stats, mc3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Special") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.4
```

```{r}
#################
#### Kmeans #####
#################
RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(1)

### Rescale the data
nba.scaled <- scale(Dat[-1], center = TRUE, scale = TRUE)
nba.scaled.attr <- attributes(nba.scaled)

km.list <- list()
wss <- NULL
for (k in 1:10){
  tmp.wss <- NULL
  ## Running kmeans 10 times for each k to get the best result
  for (itr in 1:10){
    set.seed(itr)
    km <- kmeans(nba.scaled, k)
    tmp.wss <- c(tmp.wss, km$tot.withinss)
  }
  select.seed <- which.min(tmp.wss)
  set.seed(select.seed)
  km <- kmeans(nba.scaled, k)
  wss <- c(wss, km$tot.withinss)
  km.list[[k]] <- c(km.list, km)
}

plot(seq(1,10,1), wss, type = "l", xlab = "Number of clusters")
```

```{r}
km3 <- kmeans(nba.scaled, 3) #to match nicely with mclust choose 3, can try 4, too

table(mc3$classification, km3$cluster)
```

```{r}
# Manually match the labels, as much as possible, 
# to the mclust classification for easier comparison
km3id <- ifelse(km3$cluster == 3, 2, 
                ifelse(km3$cluster == 2, 3,1))

table(km3id, Dat$Pos)
```

```{r}
table(mc3$classification, km3id)
```

```{r}
# Convert the kmeans centroids to the original scale for plotting
km3.centers.rescaled <-  km3$centers
for (j in 1:14){
  km3.centers.rescaled[,j] <- nba.scaled.attr$`scaled:center`[j] +
    km3$centers[,j]*nba.scaled.attr$`scaled:scale`[j]
}
# Now change the indices to approximately match 
# mclust classification for easier comparison
km3.centers.rescaled.tmp <- km3.centers.rescaled
km3.centers.rescaled[1,] <- km3.centers.rescaled.tmp[1,]
km3.centers.rescaled[2,] <- km3.centers.rescaled.tmp[3,]
km3.centers.rescaled[3,] <- km3.centers.rescaled.tmp[2,]

# Kmeans Clustering Plots
library(tidyr)
library(ggplot2)

dat.km <- as.data.frame(km3.centers.rescaled)
dat.km$clusters <- row.names(dat.km)

# Percentages Clusters
dat.km1 <- dat.km[c(1:5,15)]

dat.g <- gather(dat.km1, stats, km3mean, -clusters)

km.1 <- ggplot(dat.g, aes(stats, km3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Percentages") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.1
```

```{r}
# Offense Clusters
dat.km1 <- dat.km[c(6,9,14,15)]

dat.g <- gather(dat.km1, stats, km3mean, -clusters)

km.2 <- ggplot(dat.g, aes(stats, km3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Offense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.2
```

```{r}
# Defense Clusters
dat.km1 <- dat.km[c(11,10,7,15)]

dat.g <- gather(dat.km1, stats, km3mean, -clusters)

km.3 <- ggplot(dat.g, aes(stats, km3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Defense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.3
```

```{r}
# Special Clusters
dat.km1 <- dat.km[c(12,13,15)]

dat.g <- gather(dat.km1, stats, km3mean, -clusters)

km.4 <- ggplot(dat.g, aes(stats, km3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Special") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.4
```

```{r}
#### Hierarchical #####
### Create distance matrix
d <- dist(nba.scaled, method = "euclidean") 

hc.complete <- hclust(d, method="ward.D2")
plot(hc.complete) # display dendogram
hc.complete.id <- cutree(hc.complete, k=3) # cut tree into 3 clusters
rect.hclust(hc.complete, k = 3, border = "red")
```

```{r}
#dev.copy2pdf(file = "E:/Data mining/Lecture Notes/plots/dend1.pdf")

nba.hier <- Dat[-1]
nba.hier$clusters <- as.data.frame(hc.complete.id)

temp1 <- subset(nba.hier,nba.hier$clusters =='1')
temp2 <- subset(nba.hier,nba.hier$clusters =='2')
temp3 <- subset(nba.hier,nba.hier$clusters =='3')

a <- colMeans(temp1)
b <- colMeans(temp2)
c <- colMeans(temp3)

dat.hier <- rbind(a,b,c)
dat.hier <- as.data.frame(dat.hier)
dat.hier$clusters <- as.character(dat.hier$clusters)

# Hierarchical Clustering Plots
library(tidyr)
library(ggplot2)

# Percentages Clusters
dat.hier1 <- dat.hier[c(1:5,15)]

dat.g <- gather(dat.hier1, stats, hier3mean, -clusters)

hier.1 <- ggplot(dat.g, aes(stats, hier3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Percentages") +
  xlab(NULL) +
  ylab(NULL)

hier.1
```

```{r}
# Offense Clusters
dat.hier1 <- dat.hier[c(6,9,14,15)]

dat.g <- gather(dat.hier1, stats, hier3mean, -clusters)

hier.2 <- ggplot(dat.g, aes(stats, hier3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Offense") +
  xlab(NULL) +
  ylab(NULL)

hier.2
```

```{r}
# Defense Clusters
dat.hier1 <- dat.hier[c(11,10,7,15)]

dat.g <- gather(dat.hier1, stats, hier3mean, -clusters)

hier.3 <- ggplot(dat.g, aes(stats, hier3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Defense") +
  xlab(NULL) +
  ylab(NULL)

hier.3
```

```{r}
# Special Clusters
dat.hier1 <- dat.hier[c(12,13,15)]

dat.g <- gather(dat.hier1, stats, hier3mean, -clusters)

hier.4 <- ggplot(dat.g, aes(stats, hier3mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Special") +
  xlab(NULL) +
  ylab(NULL)

hier.4
```

```{r}
# Table analysis
t <- table(hc.complete.id, km3id)
rownames(t) <- c("H 1", "H 2", "H 3")
colnames(t) <- c("KM 1", "KM 2", "KM 3")

t2 <- table(mc3$classification, hc.complete.id)
rownames(t2) <- c("MC 1", "MC 2", "MC 3")
colnames(t2) <- c("H 1", "H 2", "H 3")


t3 <- table(km3id, mc3$classification)
rownames(t3) <- c("KM 1", "KM 2", "KM 3")
colnames(t3) <- c("MC 1", "MC 2", "MC 3")

t
```

```{r}
t2
```

```{r}
t3
```

```{r}
nba.clust <- cbind(Dat2, mc3$classification, km3id,hc.complete.id)
colnames(nba.clust)[17] <- 'MCid'
colnames(nba.clust)[18] <- 'KMid'
colnames(nba.clust)[19] <- 'Hierid'

nba.clust$MCid <- as.factor(nba.clust$MCid)
nba.clust$KMid <- as.factor(nba.clust$KMid)
nba.clust$Hierid <- as.factor(nba.clust$Hierid)

nba.clust
```

```{r}
################### CSV SAVED ON APRIL 9th 2021 ####################
#write.csv(nba.clust, "nba.clust3.csv")

# Compare Mcluster, Kmeans, and Hierarchical Clusters
library(gridExtra)
library(cowplot)

#par(mfrow = c(2,1))
plot_grid(mc.1, km.1, hier.1, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
plot_grid(mc.2, km.2, hier.2, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
plot_grid(mc.3, km.3, hier.3, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
plot_grid(mc.4, km.4, hier.4, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
#grid.arrange(mc.1, km.1, hier.1, nrow=1, ncol=3)
#grid.arrange(mc.2, km.2, hier.2, nrow=1, ncol=3)
#grid.arrange(mc.3, km.3, hier.3, nrow=1, ncol=3)
#grid.arrange(mc.4, km.4, hier.4, nrow=1, ncol=3)

# Clusters and Positions
nbatoday <- nba.clust[,c(1,2,17,18,19)]

dat.g <- gather(nbatoday[-1], type, cluster,-Pos)

today <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("2020 NBA Cluster Distributions: Percentage Dataset")

posclus.mc <- ggplot(nbatoday[-1], aes(Pos)) + 
  geom_bar(aes(fill = MCid), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster Relation Mclust")

posclus.km <- ggplot(nbatoday[-1], aes(Pos)) + 
  geom_bar(aes(fill = KMid), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster Relation Kmeans")

posclus.h <- ggplot(nbatoday[-1], aes(Pos)) + 
  geom_bar(aes(fill = Hierid), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster Relation Hierarchical")

posclus <- ggplot(dat.g, aes(Pos)) + 
  geom_bar(aes(fill = cluster), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster: Percentage Dataset")

today
```

```{r}
posclus.mc
```

```{r}
table(mc3$classification, Dat$Pos)
```

```{r}
posclus.km
```

```{r}
table(km3id, Dat$Pos)
```

```{r}
posclus.h
```

```{r}
table(hc.complete.id, Dat$Pos)
```

```{r}
posclus
```

```{r}
# Pairs plot view clusters
nba.nolbl <- nba.clust[,-c(1,2,17,18,19)]
nba.lbl.mc <- nba.clust[,17]
nba.lbl.km <- nba.clust[,18]
nba.lbl.h <- nba.clust[,19]
n <- nrow(nba.nolbl)
p <- ncol(nba.nolbl)

col.seq <- c("red", "darkgreen", "navy")
species.col.mc <- col.seq[as.numeric(nba.lbl.mc)]
species.col.km <- col.seq[as.numeric(nba.lbl.km)]
species.col.h <- col.seq[as.numeric(nba.lbl.h)]

# Mclust
par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[1:5], col = species.col.mc,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Model Based Cluster",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.mc)),
       fill = unique(species.col.mc))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(6,9,14,12)], col = species.col.mc,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Model Based Cluster",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.mc)),
       fill = unique(species.col.mc))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(11,10,7,13)], col = species.col.mc,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Model Based Cluster",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.mc)),
       fill = unique(species.col.mc))
```

```{r}
par(xpd = NA)

# Kmeans
par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[1:5], col = species.col.km,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Kmeans Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.km)),
       fill = unique(species.col.km))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(6,9,14,12)], col = species.col.km,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Kmeans Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.km)),
       fill = unique(species.col.km))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(11,10,7,13)], col = species.col.km,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Kmeans Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.km)),
       fill = unique(species.col.km))
```

```{r}
par(xpd = NA)

# Hierarchical
par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[1:5], col = species.col.h,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Hierarchical Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.h)),
       fill = unique(species.col.h))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(6,9,14,12)], col = species.col.h,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Hierarchical Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.h)),
       fill = unique(species.col.h))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(11,10,7,13)], col = species.col.h,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Hierarchical Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.h)),
       fill = unique(species.col.h))
```

```{r}
par(xpd = NA)

# Help us define Cluster with background knowledge of NBA (e.g. height, strength, weight, etc.)
MCid1 <- nba.clust[(nba.clust$MCid=="1"),]
MCid2 <- nba.clust[(nba.clust$MCid=="2"),]
MCid3 <- nba.clust[(nba.clust$MCid=="3"),]

MCid1
```

```{r}
MCid2
```

```{r}
MCid3
```

```{r}
KMid1 <- nba.clust[(nba.clust$KMid=="1"),]
KMid2 <- nba.clust[(nba.clust$KMid=="2"),]
KMid3 <- nba.clust[(nba.clust$KMid=="3"),]

KMid1
```

```{r}
KMid2
```

```{r}
KMid3
```

```{r}
Hid1 <- nba.clust[(nba.clust$Hierid=="1"),]
Hid2 <- nba.clust[(nba.clust$Hierid=="2"),]
Hid3 <- nba.clust[(nba.clust$Hierid=="3"),]

Hid1
```

```{r}
Hid2
```

```{r}
Hid3
```

# ATTEMPTS AND MADE DATASET

```{r}
rm(list=ls())

### Initialize data sets to work with
# Webscrape Statistics from Baskballreference
#library(rvest)
#thelink <- xml2::read_html("https://www.basketball-reference.com/leagues/NBA_2021_per_game.html")
#nba <- as.data.frame(thelink %>%html_nodes("table")%>%.[[1]] %>%html_table())

# Remove repeated header row
#unn <- subset(nba, nba$Rk != 'Rk')

# Remove duplicated player records (duplicated if player traded, keep only 'Total' record)
#nba2 <- unn[!duplicated(unn$Player),]

# Convert regressor variables to integer values
#i <- c(6:30)
#nba2[ , i] <- apply(nba2[ , i], 2, function(x) as.numeric(as.character(x)))

# Filter out players not having played >= [pct] of total games played
#pct <- .7 #define percentage of games player must have played in as compared to max player
#nba3 <- subset(nba2,nba2$G >= max(nba2$G)*pct)

# Remove attempts and attempts made columns (as they are simple divisors and dividends of percentage columns)
#nba4 <- nba3[-c(1,2,4:8,11,14,17,18,21)]
#summary(nba4)

################### CSV SAVED ON APRIL 8th 2021 ####################
#write.csv(nba3, "nba3.csv")

#nba3 <- read.csv('C:/Users/super/Documents/MSA/Sp21/MSA 6440/Final Project/nba3.csv')
nba3 <-read.csv('/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/6440 project/Dataset/nba3.csv')
colnames(nba3)[11] <- 'FG%'
colnames(nba3)[12] <- '3P'
colnames(nba3)[13] <- '3PA'
colnames(nba3)[14] <- '3P%'
colnames(nba3)[15] <- '2P'
colnames(nba3)[16] <- '2PA'
colnames(nba3)[17] <- '2P%'
colnames(nba3)[18] <- 'eF%'
colnames(nba3)[21] <- 'FT%'

nba4 <- nba3[-c(1,2,4:8,11,14,17,18,21)]

# Construct position specific dataframes and assign values to specific positions
table(nba4$Pos)
```

```{r}
nba3[nba3$Pos =='C-PF',] #Kelly Olynyk
```

```{r}
nba3[nba3$Pos =='PG-SG',] #James Harden
```

```{r}
nba3[nba3$Pos =='SF-SG',] #Sviatoslav Mykhailiuk
```

```{r}
nba3[nba3$Pos =='SG-PG',] # Delon Wright
```

```{r}
nba3[nba3$Pos =='SG-SF',] # Hamidou Diallo & Norman Powell
```

```{r}
################# KNN TO DEFINE PLAYERS ################# 
#### Olynyk: C OR PF?? ####
source("/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/R_codes/myfunctions.R")


knn <- subset(nba4,nba4$Pos =='C' | nba4$Pos =='PF')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
library(FNN)
library(caret)

K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "C")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="C-PF"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #Label as PF
```

```{r}
#### HARDIN: PG OR SG?? ####
knn <- subset(nba4,nba4$Pos =='PG' | nba4$Pos =='SG')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "PG")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="PG-SG"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label as PG
```

```{r}
#### Mykhailiuk : SF OR SG?? ####
knn <- subset(nba4,nba4$Pos =='SF' | nba4$Pos =='SG')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "SF")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="SF-SG"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label as SG
```

```{r}
#### Wright: SG OR PG?? ####
knn <- subset(nba4,nba4$Pos =='SG' | nba4$Pos =='PG')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "SG")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="SG-PG"),]
x0 <- x[-1]

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn <- knn(train = training.scaled, test = x0.scaled,
           cl = training.data[,1], k = max.k)
Knn
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label as PG / SG*
```

```{r}
table(knn$Pos) #majority rule, label SG
```

```{r}
#### Diallo & Powell: SG OR SF?? ####
knn <- subset(nba4,nba4$Pos =='SG' | nba4$Pos =='SF')

knn$Pos <- as.factor(knn$Pos)

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knn, 0.7)
training.data <- p2$data.train
test.data <- p2$data.test

### Rescale the data
training.scaled <- scale(training.data[,-1], center = TRUE, scale = TRUE)
training.scaled.wY <- cbind(training.scaled, training.data[,1])
training.scaled.attr <- attributes(training.scaled)

test.scaled <- scale(test.data[,-1], 
                     center = training.scaled.attr$`scaled:center`, 
                     scale = training.scaled.attr$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  Knn <- knn(train = training.scaled, test = test.scaled,
             cl = training.data[,1], k = kk)
  c <- confusionMatrix(as.factor(Knn), as.factor(test.data[,1]), 
                       positive = "SG")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.k <- which.max(kappa)
max.k
```

```{r}
### Fit kNN model on a single new observation with k=[max.k] 
x <- nba4[(nba4$Pos=="SG-SF"),]
x0 <- x[1,][-1] #Diallo
x1 <- x[2,][-1] #Powell

x0.scaled <- scale(x0, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 
x1.scaled <- scale(x1, center = training.scaled.attr$`scaled:center`, scale = training.scaled.attr$`scaled:scale`) 

Knn0 <- knn(train = training.scaled, test = x0.scaled,
            cl = training.data[,1], k = max.k)
Knn0
```

```{r}
Knn1 <- knn(train = training.scaled, test = x1.scaled,
            cl = training.data[,1], k = max.k)
Knn1
```

```{r}
## labels of nearest neighbors
Knn.attr <- attributes(Knn0)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label Diallo as SG MAJORITY RULE
```

```{r}
Knn.attr <- attributes(Knn1)
training.data[Knn.attr$nn.index,1]
```

```{r}
summary(training.data[Knn.attr$nn.index,1]) #label Powell as SG
```

```{r}
table(knn$Pos)
```

```{r}
#################### POP OUT OF KNN CLASSIFICIATION ###################
#define Olynyk as KNN result, viz. "PF"
nba4$Pos[nba4$Pos=="C-PF"] <- "PF"
nba3$Pos[nba3$Pos=="C-PF"] <- "PF"

#define Hardin as KNN result, viz. "PG"
nba4$Pos[nba4$Pos=="PG-SG"] <- "PG"
nba3$Pos[nba3$Pos=="PG-SG"] <- "PG"

#define Mykhailiuk  as KNN result, viz. "SG"
nba4$Pos[nba4$Pos=="SF-SG"] <- "SG"
nba3$Pos[nba3$Pos=="SF-SG"] <- "SG"

#define Write as KNN result, viz. "SG"
nba4$Pos[nba4$Pos=="SG-PG"] <- "SG"
nba3$Pos[nba3$Pos=="SG-PG"] <- "SG"

#define Powell as KNN result, viz. "SG"
nba4$Pos[nba4$Pos=="SG-SF"] <- "SG"
nba3$Pos[nba3$Pos=="SG-SF"] <- "SG"

Dat <- nba4
Dat2 <- nba3[-c(1,4:8,11,14,17,18,21)]

Dat$Pos <- as.factor(Dat$Pos)
Dat2$Pos <- as.factor(Dat2$Pos)

################### CSV SAVED ON APRIL 11th 2021 ####################
#write.csv(Dat2, "NBAppa.csv")

## Create training and test data ##
#source("C:/Users/super/Documents/MSA/Sp21/MSA 6440/Data/myfunctions.R")
source("/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/R_codes/myfunctions.R")

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0) ## set seed so that you get same partition each time
p2 <- partition.2(Dat, 0.7) ## creating 70:30 partition
training.data <- p2$data.train
test.data <- p2$data.test

nba.nolbl <- training.data[,-1]
nba.scaled <- scale(nba.nolbl, center = TRUE, scale = TRUE)

# https://predictivehacks.com/how-to-determine-the-number-of-clusters-of-k-means-in-r/
library(factoextra)
library(NbClust)
library(mclust)

set.seed(123)

fviz_nbclust(nba.scaled, kmeans, method = "wss") +
  geom_vline(xintercept = 2, linetype = 2)+
  labs(subtitle = "Elbow method")
```

```{r}
fviz_nbclust(nba.scaled, kmeans, method = "silhouette")+
  labs(subtitle = "Silhouette method")
```

```{r}
fviz_nbclust(nba.scaled, kmeans, nstart = 25,  method = "gap_stat", nboot = 500)+
  labs(subtitle = "Gap statistic method") # Let's double check with Mclust
```

```{r}
############# Clustering ##################
#### Mclust #####
library(mclust)
RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)

mc.list <- list()
bic <- NULL
for (k in 1:10){
  mc <- Mclust(Dat[-1], k)
  bic <- c(bic, mc$bic)
  mc.list[[k]] <- c(mc.list, mc)
}

plot(seq(1,10,1), bic, type = "l", xlab = "Number of clusters")
```

```{r}
plot(bic[1:10], type = "l", xlab = "Number of clusters")
```

```{r}
maxbic <- which.max(bic[1:10])
maxbic # cross-checked with gap-statistic clustering
```

```{r}
mc4 <- Mclust(Dat[-1], maxbic)

table(mc4$classification, Dat$Pos)
```

```{r}
# Mclust Clustering Plots
library(tidyr)
library(ggplot2)

dat.mc <- as.data.frame(t(mc4$parameters$mean))
dat.mc$clusters <- row.names(dat.mc)

# Shots Clusters
dat.mc1 <- dat.mc[c(1:8,18)]

dat.g <- gather(dat.mc1, stats, mc4mean, -clusters)

mc.1 <- ggplot(dat.g, aes(stats, mc4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Shots") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.1
```

```{r}
# Offense Clusters
dat.mc1 <- dat.mc[c(9,12,17,18)]

dat.g <- gather(dat.mc1, stats, mc4mean, -clusters)

mc.2 <- ggplot(dat.g, aes(stats, mc4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Offense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.2
```

```{r}
# Defense Clusters
dat.mc1 <- dat.mc[c(14,13,10,18)]

dat.g <- gather(dat.mc1, stats, mc4mean, -clusters)

mc.3 <- ggplot(dat.g, aes(stats, mc4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Defense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.3
```

```{r}
# Special Clusters
dat.mc1 <- dat.mc[c(15,16,18)]

dat.g <- gather(dat.mc1, stats, mc4mean, -clusters)

mc.4 <- ggplot(dat.g, aes(stats, mc4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nMclust: Special") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

mc.4
```

```{r}
#### Kmeans #####
RNGkind (sample.kind = "Rounding")
```

```{r}
set.seed(1)

### Rescale the data
nba.scaled <- scale(Dat[-1], center = TRUE, scale = TRUE)
nba.scaled.attr <- attributes(nba.scaled)

km.list <- list()
wss <- NULL
for (k in 1:10){
  tmp.wss <- NULL
  ## Running kmeans 10 times for each k to get the best result
  for (itr in 1:10){
    set.seed(itr)
    km <- kmeans(nba.scaled, k)
    tmp.wss <- c(tmp.wss, km$tot.withinss)
  }
  select.seed <- which.min(tmp.wss)
  set.seed(select.seed)
  km <- kmeans(nba.scaled, k)
  wss <- c(wss, km$tot.withinss)
  km.list[[k]] <- c(km.list, km)
}

plot(seq(1,10,1), wss, type = "l", xlab = "Number of clusters")
```

```{r}
km4 <- kmeans(nba.scaled, 4)

table(mc4$classification, km4$cluster)
```

```{r}
# Manually match the labels, as much as possible, 
# to the mclust classification for easier comparison
km4id <- ifelse(km4$cluster == 4, 1, 
                ifelse(km4$cluster == 3, 2,
                       ifelse(km4$cluster == 2, 4, 3)))

table(km4id, Dat$Pos) 
```

```{r}
table(mc4$classification, km4id)
```

```{r}
# Convert the kmeans centroids to the original scale for plotting
km4.centers.rescaled <-  km4$centers
for (j in 1:17){
  km4.centers.rescaled[,j] <- nba.scaled.attr$`scaled:center`[j] +
    km4$centers[,j]*nba.scaled.attr$`scaled:scale`[j]
}
# Now change the indices to approximately match 
# mclust classification for easier comparison
km4.centers.rescaled.tmp <- km4.centers.rescaled
km4.centers.rescaled[1,] <- km4.centers.rescaled.tmp[4,]
km4.centers.rescaled[2,] <- km4.centers.rescaled.tmp[3,]
km4.centers.rescaled[3,] <- km4.centers.rescaled.tmp[1,]
km4.centers.rescaled[4,] <- km4.centers.rescaled.tmp[2,]

# Kmeans Clustering Plots
library(tidyr)
library(ggplot2)

dat.km <- as.data.frame(km4.centers.rescaled)
dat.km$clusters <- row.names(dat.km)

# Shots Clusters
dat.km1 <- dat.km[c(1:8,18)]

dat.g <- gather(dat.km1, stats, km4mean, -clusters)

km.1 <- ggplot(dat.g, aes(stats, km4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Shots") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.1
```

```{r}
# Offense Clusters
dat.km1 <- dat.km[c(9,12,17,18)]

dat.g <- gather(dat.km1, stats, km4mean, -clusters)

km.2 <- ggplot(dat.g, aes(stats, km4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Offense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.2
```

```{r}
# Defense Clusters
dat.km1 <- dat.km[c(14,13,10,18)]

dat.g <- gather(dat.km1, stats, km4mean, -clusters)

km.3 <- ggplot(dat.g, aes(stats, km4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Defense") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.3
```

```{r}
# Special Clusters
dat.km1 <- dat.km[c(15,16,18)]

dat.g <- gather(dat.km1, stats, km4mean, -clusters)

km.4 <- ggplot(dat.g, aes(stats, km4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nKmeans: Special") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

km.4
```

```{r}
#### Hierarchical #####
### Create distance matrix
d <- dist(nba.scaled, method = "euclidean") 

hc.complete <- hclust(d, method="ward.D2")
plot(hc.complete) # display dendogram
hc.complete.id <- cutree(hc.complete, k=4) # cut tree into 3 clusters
rect.hclust(hc.complete, k = 4, border = "red")
```

```{r}
#dev.copy2pdf(file = "E:/Data mining/Lecture Notes/plots/dend1.pdf")

nba.hier <- Dat[-1]
nba.hier$clusters <- as.data.frame(hc.complete.id)

temp1 <- subset(nba.hier,nba.hier$clusters =='1')
temp2 <- subset(nba.hier,nba.hier$clusters =='2')
temp3 <- subset(nba.hier,nba.hier$clusters =='3')
temp4 <- subset(nba.hier,nba.hier$clusters =='4')

a <- colMeans(temp1)
b <- colMeans(temp2)
c <- colMeans(temp3)
d <- colMeans(temp4)

dat.hier <- rbind(a,b,c,d)
dat.hier <- as.data.frame(dat.hier)
dat.hier$clusters <- as.character(dat.hier$clusters)

# Hierarchical Clustering Plots
library(tidyr)
library(ggplot2)

# Shots Clusters
dat.hier1 <- dat.hier[c(1:8,18)]

dat.g <- gather(dat.hier1, stats, hier4mean, -clusters)

hier.1 <- ggplot(dat.g, aes(stats, hier4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Shots") +
  xlab(NULL) +
  ylab(NULL)

hier.1
```

```{r}
# Offense Clusters
dat.hier1 <- dat.hier[c(9,12,17,18)]

dat.g <- gather(dat.hier1, stats, hier4mean, -clusters)

hier.2 <- ggplot(dat.g, aes(stats, hier4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Offense") +
  xlab(NULL) +
  ylab(NULL)

hier.2
```

```{r}
# Defense Clusters
dat.hier1 <- dat.hier[c(14,13,10,18)]

dat.g <- gather(dat.hier1, stats, hier4mean, -clusters)

hier.3 <- ggplot(dat.g, aes(stats, hier4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Defense") +
  xlab(NULL) +
  ylab(NULL)

hier.3
```

```{r}
# Special Clusters
dat.hier1 <- dat.hier[c(15,16,18)]

dat.g <- gather(dat.hier1, stats, hier4mean, -clusters)

hier.4 <- ggplot(dat.g, aes(stats, hier4mean)) + 
  geom_bar(aes(fill = clusters), stat = "identity", position = "dodge") +
  ggtitle("Clustering using \nHierarchical: Special") +
  xlab(NULL) +
  ylab(NULL)

hier.4
```

```{r}
# Table analysis
t <- table(km4id, hc.complete.id)
colnames(t) <- c("H 1", "H 2", "H 3", "H 4")
rownames(t) <- c("KM 1", "KM 2", "KM 3", "KM 4")

t2 <- table(mc4$classification, hc.complete.id) 
colnames(t2) <- c("H 1", "H 2", "H 3", "H 4")
rownames(t2) <- c("MC 1", "MC 2", "MC 3", "MC 4")

table(mc4$classification, km4id)
```

```{r}
t
```

```{r}
t2
```

```{r}
nba.clust <- cbind(Dat2, mc4$classification, km4id,hc.complete.id)
colnames(nba.clust)[20] <- 'MCid'
colnames(nba.clust)[21] <- 'KMid'
colnames(nba.clust)[22] <- 'Hierid'

nba.clust$MCid <- as.factor(nba.clust$MCid)
nba.clust$KMid <- as.factor(nba.clust$KMid)
nba.clust$Hierid <- as.factor(nba.clust$Hierid)

nba.clust
```

```{r}
################### CSV SAVED ON APRIL 8th 2021 ####################
#write.csv(nba.clust, "nba.clust4.csv")

# Compare Mcluster, Kmeans, and Hierarchical Clusters
library(gridExtra)
library(cowplot)

#par(mfrow = c(2,1))
plot_grid(mc.1, km.1, hier.1, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
plot_grid(mc.2, km.2, hier.2, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
plot_grid(mc.3, km.3, hier.3, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
plot_grid(mc.4, km.4, hier.4, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```

```{r}
#grid.arrange(mc.1, km.1, hier.1, nrow=1, ncol=3)
#grid.arrange(mc.2, km.2, hier.2, nrow=1, ncol=3)
#grid.arrange(mc.3, km.3, hier.3, nrow=1, ncol=3)
#grid.arrange(mc.4, km.4, hier.4, nrow=1, ncol=3)

# Clusters and Positions
nbatoday <- nba.clust[,c(1,2,20,21,22)]

dat.g <- gather(nbatoday[-1], type, cluster,-Pos)

today <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("2020 NBA Cluster Distributions: A&M Dataset")

posclus.mc <- ggplot(nbatoday[-1], aes(Pos)) + 
  geom_bar(aes(fill = MCid), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster Relation Mclust")

posclus.km <- ggplot(nbatoday[-1], aes(Pos)) + 
  geom_bar(aes(fill = KMid), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster Relation Kmeans")

posclus.h <- ggplot(nbatoday[-1], aes(Pos)) + 
  geom_bar(aes(fill = Hierid), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster Relation Hierarchical")

posclus <- ggplot(dat.g, aes(Pos)) + 
  geom_bar(aes(fill = cluster), stat = "count", position = "dodge") +
  ggtitle("2020 NBA Position-Cluster: A&M Dataset")

today
```

```{r}
posclus.mc
```

```{r}
table(mc4$classification, Dat$Pos)
```

```{r}
posclus.km
```

```{r}
table(km4id, Dat$Pos)
```

```{r}
posclus.h
```

```{r}
table(hc.complete.id, Dat$Pos)
```

```{r}
posclus
```

```{r}
# Pairs plot view clusters
nba.nolbl <- nba.clust[,-c(1,2,20,21,22)]
nba.lbl.mc <- nba.clust[,20]
nba.lbl.km <- nba.clust[,21]
nba.lbl.h <- nba.clust[,22]
n <- nrow(nba.nolbl)
p <- ncol(nba.nolbl)

col.seq <- c("red", "darkgreen", "lightblue", "purple")
species.col.mc <- col.seq[as.numeric(nba.lbl.mc)]
species.col.km <- col.seq[as.numeric(nba.lbl.km)]
species.col.h <- col.seq[as.numeric(nba.lbl.h)]

# Mclust
par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[1:8], col = species.col.mc,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Model Based Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.mc)),
       fill = unique(species.col.mc))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(9,12,17,15)], col = species.col.mc,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Model Based Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.mc)),
       fill = unique(species.col.mc))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(14,13,10,16)], col = species.col.mc,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Model Based Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.mc)),
       fill = unique(species.col.mc))
```

```{r}
par(xpd = NA)

# Kmeans
par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[1:8], col = species.col.km,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Kmeans Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.km)),
       fill = unique(species.col.km))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(9,12,17,15)], col = species.col.km,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Kmeans Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.km)),
       fill = unique(species.col.km))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(14,13,10,16)], col = species.col.km,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Kmeans Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.km)),
       fill = unique(species.col.km))

```

```{r}
par(xpd = NA)

# Hierarchical
par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[1:8], col = species.col.h,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Hierarchical Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.h)),
       fill = unique(species.col.h))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(9,12,17,15)], col = species.col.h,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Hierarchical Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.h)),
       fill = unique(species.col.h))
```

```{r}
par(xpd = NA)

par(oma=c(2,2,4,1), mar = c(4,4,4,1) + 0.1)
pairs(nba.nolbl[c(14,13,10,16)], col = species.col.h,
      lower.panel = NULL, cex.labels=2, pch=19, cex = 1.2)
mtext(text= "Hierarchical Clusters",side=3,line=3, font=2,
      cex = 1.25,outer=TRUE)
legend(x = 0.005, y = 0.3, cex = 1.2, bty = "n",
       legend = as.character(levels(nba.lbl.h)),
       fill = unique(species.col.h))
```

```{r}
par(xpd = NA)

# Help us define Cluster with background knowledge of NBA (e.g. height, strength, weight, etc.)
MCid1 <- nba.clust[(nba.clust$MCid=="1"),]
MCid2 <- nba.clust[(nba.clust$MCid=="2"),]
MCid3 <- nba.clust[(nba.clust$MCid=="3"),]
MCid4 <- nba.clust[(nba.clust$MCid=="4"),]

MCid1
```

```{r}
MCid2
```

```{r}
MCid3
```

```{r}
MCid4
```

```{r}
KMid1 <- nba.clust[(nba.clust$KMid=="1"),]
KMid2 <- nba.clust[(nba.clust$KMid=="2"),]
KMid3 <- nba.clust[(nba.clust$KMid=="3"),]
KMid4 <- nba.clust[(nba.clust$KMid=="4"),]

KMid1
```

```{r}
KMid2
```

```{r}
KMid3
```

```{r}
KMid4
```

```{r}
Hid1 <- nba.clust[(nba.clust$Hierid=="1"),]
Hid2 <- nba.clust[(nba.clust$Hierid=="2"),]
Hid3 <- nba.clust[(nba.clust$Hierid=="3"),]
Hid4 <- nba.clust[(nba.clust$Hierid=="4"),]

Hid1
```

```{r}
Hid2
```

```{r}
Hid3
```

```{r}
Hid4
```

# EXTENSION NBA CHAMPIONSHIP

```{r}
rm(list=ls())
library(rvest)

nba.clust.x <- read.csv('/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/6440 project/Dataset/nba.clust3.csv')
  #read.csv('C:/Users/super/Documents/MSA/Sp21/MSA 6440/Final Project/nba.clust3.csv')
colnames(nba.clust.x)[3] <- 'FG%'
colnames(nba.clust.x)[4] <- '3P%'
colnames(nba.clust.x)[5] <- '2P%'
colnames(nba.clust.x)[6] <- 'eF%'
colnames(nba.clust.x)[7] <- 'FT%'

nba.clust.x$MCid <- as.factor(nba.clust.x$MCid)
nba.clust.x$KMid <- as.factor(nba.clust.x$KMid)
nba.clust.x$Hierid <- as.factor(nba.clust.x$Hierid)

nba.clust <- nba.clust.x[-2]

####### Post Analysis Interesting Question: Champion Clusters ##############
####### 2020 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/2020.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
#curl::curl("https://www.basketball-reference.com/teams/LAL/2020.html") %>% read_html()
```


```{r}
dat4 <- dat3[-1]

#### Lakers categorizations McCluster####
knnMC <- nba.clust[-c(1,17,18)]

RNGkind (sample.kind = "Rounding") 
```

```{r}
source("/Users/shegeb/Documents/BGSU/Spring 2021/data_mining/R_codes/myfunctions.R")
set.seed(0)

#source("C:/Users/super/Documents/MSA/Sp21/MSA 6440/Data/myfunctions.R")

### create 70:30 partition
p2 <- partition.2(knnMC, 0.7)
training.data.mc <- p2$data.train
test.data.mc <- p2$data.test

### Rescale the data
training.scaled.mc <- scale(training.data.mc[,-15], center = TRUE, scale = TRUE)
training.scaled.wY.mc <- cbind(training.scaled.mc, training.data.mc[,15])
training.scaled.attr.mc <- attributes(training.scaled.mc)

test.scaled.mc <- scale(test.data.mc[,-15], 
                        center = training.scaled.attr.mc$`scaled:center`, 
                        scale = training.scaled.attr.mc$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
library(FNN)
library(caret)

K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  knnMC <- knn(train = training.scaled.mc, test = test.scaled.mc,
               cl = training.data.mc[,15], k = kk)
  c <- confusionMatrix(as.factor(knnMC), as.factor(test.data.mc[,15]), 
                       positive = "C")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.kmc <- which.max(kappa)
max.kmc
```

```{r}
### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
######Lakers categorizations Kmeans#########
knnKM <- nba.clust[-c(1,16,18)]

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knnKM, 0.7)
training.data.km <- p2$data.train
test.data.km <- p2$data.test

### Rescale the data
training.scaled.km <- scale(training.data.km[,-15], center = TRUE, scale = TRUE)
training.scaled.wY.km <- cbind(training.scaled.km, training.data.km[,15])
training.scaled.attr.km <- attributes(training.scaled.km)

test.scaled.km <- scale(test.data.km[,-15], 
                        center = training.scaled.attr.km$`scaled:center`, 
                        scale = training.scaled.attr.km$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  knnMC <- knn(train = training.scaled.km, test = test.scaled.km,
               cl = training.data.km[,15], k = kk)
  c <- confusionMatrix(as.factor(knnMC), as.factor(test.data.km[,15]), 
                       positive = "C")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.kkm <- which.max(kappa)
max.kkm
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

###### categorizations Hierarchical #########
knnH <- nba.clust[-c(1,16,17)]

RNGkind (sample.kind = "Rounding") 
```

```{r}
set.seed(0)
### create 70:30 partition
p2 <- partition.2(knnH, 0.7)
training.data.h <- p2$data.train
test.data.h <- p2$data.test

### Rescale the data
training.scaled.h <- scale(training.data.h[,-15], center = TRUE, scale = TRUE)
training.scaled.wY.h <- cbind(training.scaled.h, training.data.h[,15])
training.scaled.attr.h <- attributes(training.scaled.h)

test.scaled.h <- scale(test.data.h[,-15], 
                       center = training.scaled.attr.h$`scaled:center`, 
                       scale = training.scaled.attr.h$`scaled:scale`)

### fit k-nn model for k = 1, ..., 60
K <- 60
kappa <- rep(0, K)
for (kk in 1:K){
  knnMC <- knn(train = training.scaled.h, test = test.scaled.h,
               cl = training.data.h[,15], k = kk)
  c <- confusionMatrix(as.factor(knnMC), as.factor(test.data.h[,15]), 
                       positive = "C")
  kappa[kk] <- c$overall["Kappa"]
  cat("K", kk, "Kappa", kappa[kk], "\n")
}
```

```{r}
# create a plot for k vs. kappa
plot(c(1:K), kappa, xlab = "k", ylab = "Kappa", type = "l", col = "blue")
```

```{r}
max.h <- which.max(kappa)
max.h
```

```{r}
max.h <- 3 #1 too low and 3 has same kappa

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers2020 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers2020)[4] <- 'Hierid'

lakers2020
```

```{r}
####### 2019 Champions: Toronto Raptors ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/TOR/2019.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

raptors2019 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(raptors2019)[4] <- 'Hierid'

raptors2019
```

```{r}
####### 2018 Champions: Golden State Warriors ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/GSW/2018.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

warriors2018 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(warriors2018)[4] <- 'Hierid'

warriors2018
```

```{r}
####### 2017 Champions: Golden State Warriors ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/GSW/2017.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

warriors2017 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(warriors2017)[4] <- 'Hierid'

warriors2017
```

```{r}
####### 2016 Champions: Cleveland Cavaliers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CLE/2016.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

cavaliers2016 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(cavaliers2016)[4] <- 'Hierid'

cavaliers2016
```

```{r}
####### 2015 Champions: Golden State Warriors ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/GSW/2015.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

warriors2015 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(warriors2015)[4] <- 'Hierid'

warriors2015
```

```{r}
####### 2014 Champions: San Antonio Spurs ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/SAS/2014.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

spurs2014 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(spurs2014)[4] <- 'Hierid'

spurs2014
```

```{r}
####### 2013 Champions: Miami Heat ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/MIA/2013.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

heat2013 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(heat2013)[4] <- 'Hierid'

heat2013
```

```{r}
####### 2012 Champions: Miami Heat ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/MIA/2012.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

heat2012 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(heat2012)[4] <- 'Hierid'

heat2012
```

```{r}
####### 2011 Champions: Dallas Mavericks ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/DAL/2011.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Raptors3 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

mavericks2011 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(mavericks2011)[4] <- 'Hierid'

mavericks2011
```

```{r}
########### OLD TIMER CHAMPIONS CATEGORIZED WITH MODELED CATEGORIES ###################
####### 1980 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/1980.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers1980 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers1980)[4] <- 'Hierid'

lakers1980
```

```{r}
####### 1981 Champions: Boston Celtics ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/BOS/1981.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Celtics 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Celtics 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

celtics1981 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(celtics1981)[4] <- 'Hierid'

celtics1981
```

```{r}
####### 1982 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/1982.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers1982 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers1982)[4] <- 'Hierid'

lakers1982
```

```{r}
####### 1983 Champions: Philadelphia 76ers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/PHI/1983.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on 76ers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on 76ers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

s76ers1983 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(s76ers1983)[4] <- 'Hierid'

s76ers1983
```

```{r}
####### 1984 Champions: Boston Celtics ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/BOS/1984.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Celtics 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Celtics 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

celtics1984 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(celtics1984)[4] <- 'Hierid'

celtics1984
```

```{r}
####### 1985 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/1985.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers1985 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers1985)[4] <- 'Hierid'

lakers1985
```

```{r}
####### 1986 Champions: Boston Celtics ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/BOS/1986.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Celtics 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Celtics 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

celtics1986 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(celtics1986)[4] <- 'Hierid'

celtics1986
```

```{r}
####### 1987 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/1987.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

celtics1987 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(celtics1987)[4] <- 'Hierid'

celtics1987
```

```{r}
####### 1988 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/1988.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers1988 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers1988)[4] <- 'Hierid'

lakers1988
```

```{r}
####### 1989 Champions: Detroit Pistons ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/DET/1989.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Pistons 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Pistons 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

pistons1989 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(pistons1989)[4] <- 'Hierid'

pistons1989
```

```{r}
####### 1990 Champions: Detroit Pistons ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/DET/1990.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Pistons 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Pistons 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

pistons1990 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(pistons1990)[4] <- 'Hierid'

pistons1990
```

```{r}
####### 1991 Champions: Chicago Bulls ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CHI/1991.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Bulls 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Bulls 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

bulls1991 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(bulls1991)[4] <- 'Hierid'

bulls1991
```

```{r}
####### 1992 Champions: Chicago Bulls ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CHI/1992.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Bulls 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Bulls 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

bulls1992 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(bulls1992)[4] <- 'Hierid'

bulls1992
```

```{r}
####### 1993 Champions: Chicago Bulls ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CHI/1993.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Bulls 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Bulls 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

bulls1993 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(bulls1993)[4] <- 'Hierid'

bulls1993
```

```{r}
####### 1994 Champions: Houston Rockets ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/HOU/1994.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Rockets 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Rockets 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

rockets1994 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(rockets1994)[4] <- 'Hierid'

rockets1994
```

```{r}
####### 1995 Champions: Houston Rockets ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/HOU/1995.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Rockets 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Rockets 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

rockets1995 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(rockets1995)[4] <- 'Hierid'

rockets1995
```

```{r}
####### 1996 Champions: Chicago Bulls ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CHI/1996.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Bulls 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Bulls 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

bulls1996 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(bulls1996)[4] <- 'Hierid'

bulls1996
```

```{r}
####### 1997 Champions: Chicago Bulls ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CHI/1997.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Bulls 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Bulls 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

bulls1997 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(bulls1997)[4] <- 'Hierid'

bulls1997
```

```{r}
####### 1998 Champions: Chicago Bulls ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/CHI/1998.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Bulls 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Bulls 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

bulls1998 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(bulls1998)[4] <- 'Hierid'

bulls1998
```

```{r}
####### 1999 Champions: San Antonio Spurs ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/SAS/1999.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Spurs 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Spurs 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

spurs1999 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(spurs1999)[4] <- 'Hierid'

spurs1999
```

```{r}

```


```{r}
####### 2000 Champions: Los Angeles Lakers ###########

newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/2000.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())

#library(rvest)
#url = "https://www.basketball-reference.com/teams/LAL/2000.html"
#download.file(url, destfile = "scrapedpage.html", quiet=TRUE)
#content <- read_html("scrapedpage.html")


dat <- as.data.frame(dat %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3

dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers2000 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers2000)[4] <- 'Hierid'

lakers2000
```

```{r}
####### 2001 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/2001.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers2001 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers2001)[4] <- 'Hierid'

lakers2001
```

```{r}
####### 2002 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/2002.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers2002 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers2002)[4] <- 'Hierid'

lakers2002
```

```{r}
####### 2003 Champions: San Antonio Spurs ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/SAS/2003.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Spurs 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Spurs 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

spurs2003 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(spurs2003)[4] <- 'Hierid'

spurs2003
```

```{r}
####### 2004 Champions: Detroit Pistons ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/DET/2004.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Pistons 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Pistons 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

pistons2004 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(pistons2004)[4] <- 'Hierid'

pistons2004
```

```{r}
####### 2005 Champions: San Antonio Spurs ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/SAS/2005.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Spurs 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Spurs 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

spurs2005 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(spurs2005)[4] <- 'Hierid'

spurs2005
```

```{r}
####### 2006 Champions: Miami Heat ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/MIA/2006.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Heat 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Heat 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

heat2006 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(heat2006)[4] <- 'Hierid'

heat2006
```

```{r}
####### 2007 Champions: San Antonio Spurs ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/SAS/2007.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Spurs 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Spurs 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

spurs2007 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(spurs2007)[4] <- 'Hierid'

spurs2007
```

```{r}
####### 2008 Champions: Boston Celtics ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/BOS/2008.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Celtics 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Celtics 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

celtics2008 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(celtics2008)[4] <- 'Hierid'

celtics2008
```

```{r}
####### 2009 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/2009.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]
dat4$`3P%`[is.na(dat4$`3P%`)]<-0 

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers2009 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers2009)[4] <- 'Hierid'

lakers2009
```

```{r}
####### 2010 Champions: Los Angeles Lakers ###########
newlink <- xml2::read_html("https://www.basketball-reference.com/teams/LAL/2010.html")
dat <- as.data.frame(newlink %>%html_nodes("table")%>%.[[2]] %>%html_table())
dat2 <- subset(dat,dat$G > 0.7 * max(dat$G))
dat3 <- dat2[-c(1,3:8,10,11,13,14,17,18)]
colnames(dat3)[1] <- "Player"
colnames(dat3)[15] <- "PTS"

dat3
```

```{r}
dat4 <- dat3[-1]

### Fit kNN model on Lakers 
levels_a <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.mc$`scaled:center`, scale = training.scaled.attr.mc$`scaled:scale`)
  c <- knn(train = training.scaled.mc, test = b,
           cl = training.data.mc[,15], k = max.kmc)
  test <- levels(c)
  levels_a[i] <- test
}

data <- cbind(dat3[1], t(as.data.frame(levels_a)))
colnames(data)[2] <- 'MCid'
data
```

```{r}
### Fit kNN model on Lakers 
levels_b <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.km$`scaled:center`, scale = training.scaled.attr.km$`scaled:scale`)
  c <- knn(train = training.scaled.km, test = b,
           cl = training.data.km[,15], k = max.kkm)
  test <- levels(c)
  levels_b[i] <- test
}

data2 <- cbind(data, t(as.data.frame(levels_b)))
colnames(data2)[3] <- 'KMid'

### Fit Hierarchical model 
levels_c <- list()
for (i in c(1:nrow(dat3))){
  a <- dat4[i,]
  b <- scale(a, center = training.scaled.attr.h$`scaled:center`, scale = training.scaled.attr.h$`scaled:scale`)
  c <- knn(train = training.scaled.h, test = b,
           cl = training.data.h[,15], k = max.h)
  test <- levels(c)
  levels_c[i] <- test
}

lakers2010 <- cbind(data2, t(as.data.frame(levels_c)))
colnames(lakers2010)[4] <- 'Hierid'

lakers2010
```

```{r}
########### Row binds Champions in two types, past 5 years and 10 years ###############
nbatoday <- nba.clust[,c(1,16,17,18)]

champs80s <- rbind(lakers1980,celtics1981,lakers1982,s76ers1983,celtics1984,lakers1985,celtics1986,celtics1987,lakers1988,pistons1989)
champs90s <- rbind(pistons1990,bulls1991,bulls1992,bulls1993,rockets1994,rockets1995,bulls1996,bulls1997,bulls1998,spurs1999)
champs00s <- rbind(lakers2000,lakers2001,lakers2002,spurs2003,pistons2004,spurs2005,heat2006,spurs2007,celtics2008,lakers2009)
champs10s <- rbind(lakers2010,mavericks2011,heat2012,heat2013,spurs2014,warriors2015,cavaliers2016,warriors2017,warriors2018,raptors2019)

e1980s <- rbind(lakers1980,celtics1981,lakers1982,s76ers1983,celtics1984)
l1980s <- rbind(lakers1985,celtics1986,celtics1987,lakers1988,pistons1989)
e1990s <- rbind(pistons1990,bulls1991,bulls1992,bulls1993,rockets1994)
l1990s <- rbind(rockets1995,bulls1996,bulls1997,bulls1998,spurs1999)
e2000s <- rbind(lakers2000,lakers2001,lakers2002,spurs2003,pistons2004)
l2000s <- rbind(spurs2005,heat2006,spurs2007,celtics2008,lakers2009)
e2010s <- rbind(lakers2010,mavericks2011,heat2012,heat2013,spurs2014)
l2010s <- rbind(warriors2015,cavaliers2016,warriors2017,warriors2018,raptors2019)

library(tidyr)
library(ggplot2)

dat.g <- gather(champs80s, type, cluster,-Player)

clust80s <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("1980's Championship Teams \nCluster Distributions") +
  xlab(NULL) +
  ylab(NULL)

dat.g <- gather(champs90s, type, cluster,-Player)

clust90s <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("1990's Championship Teams \nCluster Distributions") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

dat.g <- gather(champs00s, type, cluster,-Player)

clust00s <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("2000's Championship Teams \nCluster Distributions") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

dat.g <- gather(champs10s, type, cluster,-Player)

clust10s <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("2010's Championship Teams \nCluster Distributions") +
  theme(legend.position="none") +
  xlab(NULL) +
  ylab(NULL)

dat.g <- gather(nbatoday, type, cluster,-Player)

today <- ggplot(dat.g[-1], aes(type,fill=cluster)) +
  geom_bar(position = "dodge") +
  ggtitle("NBA Today Cluster Distributions")

clust80s
```

```{r}
clust90s
```

```{r}
clust00s
```

```{r}
clust10s
```

```{r}
today
```

```{r}
library(gridExtra)
library(cowplot)

clust80s
```

```{r}
plot_grid(clust90s, clust00s, clust10s, align = "h", ncol = 3, rel_widths = c(3/10,3/10,4/10))
```