Add README with explanation

Author: ph1ps

Naive Bayes Classifier/NaiveBayes.playground/Contents.swift

@@ -2,13 +2,13 @@ import Foundation
 
 /*:
  ## Naive Bayes Classifier
- 
+
  This playground uses the given algorithm and utilizes its features with some example datasets
- 
+
  ### Gaussian Naive Bayes
  - Note:
  When using Gaussian NB you have to have continuous features (Double).
- 
+
  For this example we are going to use a famous dataset with different types of wine. The labels of the features can be viewed [here](https://gist.github.com/tijptjik/9408623)
  */
 guard let wineCSV = Bundle.main.path(forResource: "wine", ofType: "csv") else {
@@ -27,7 +27,7 @@ guard let csv = try? String(contentsOfFile: wineCSV) else {
 let rows = csv.characters.split(separator: "\r\n").map { String($0) }
 /*:
  Splitting on the ; sign and converting the value to a Double
- 
+
  - Important:
  Do not force unwrap the mapped values in your real application. Carefully convert them! This is just for the sake of showing how the algorithm works.
  */
@@ -47,31 +47,31 @@ let data = wineData.map { row in
 
 /*:
  Again use `guard` on the result of a `try?` or simply `do-try-catch` because this would crash your application if an error occured.
- 
+
  The array in the `classifyProba` method I passed is a former entry in the .csv file which I removed in order to classify it.
  */
 let wineBayes = try! NaiveBayes(type: .gaussian, data: data, classes: classes).train()
 let result = wineBayes.classifyProba(with: [12.85, 1.6, 2.52, 17.8, 95, 2.48, 2.37, 0.26, 1.46, 3.93, 1.09, 3.63, 1015])
 /*:
  I can assure you that ***class 1*** is the correct result and as you can see the classifier thinks that its ***99.99%*** likely too.
- 
+
  ### Multinomial Naive Bayes
- 
+
  - Note:
  When using Multinomial NB you have to have categorical features (Int).
- 
+
  Now this dataset is commonly used to describe the classification problem and it is categorical which means you don't have real values you just have categorical data as stated before. The structure of this dataset is as follows.
- 
+
  Outlook,Temperature,Humidity,Windy
- 
+
  ***Outlook***: 0 = rainy, 1 = overcast, 2 = sunny
- 
+
  ***Temperature***: 0 = hot, 1 = mild, 2 = cool
- 
+
  ***Humidity***: 0 = high, 1 = normal
- 
+
  ***Windy***: 0 = false, 1 = true
- 
+
  The classes are either he will play golf or not depending on the weather conditions. (0 = won't play, 1 = will play)
  */
 

Naive Bayes Classifier/NaiveBayes.playground/Sources/NaiveBayes.swift

@@ -11,186 +11,186 @@ import Foundation
 extension String: Error {}
 
 extension Array where Element == Double {
-    
+
     func mean() -> Double {
         return self.reduce(0, +) / Double(count)
     }
-    
+
     func standardDeviation() -> Double {
         let calculatedMean = mean()
-        
+
         let sum = self.reduce(0.0) { (previous, next) in
             return previous + pow(next - calculatedMean, 2)
         }
-        
+
         return sqrt(sum / Double(count - 1))
     }
 }
 
 extension Array where Element == Int {
-    
+
     func uniques() -> Set<Element> {
         return Set(self)
     }
-    
+
 }
 
 public enum NBType {
-    
+
     case gaussian
     case multinomial
     //case bernoulli --> TODO
-    
+
     func calcLikelihood(variables: [Any], input: Any) -> Double? {
-        
+
         if case .gaussian = self {
-            
+
             guard let input = input as? Double else {
                 return nil
             }
-            
+
             guard let mean = variables[0] as? Double else {
                 return nil
             }
-            
+
             guard let stDev = variables[1] as? Double else {
                 return nil
             }
-            
+
             let eulerPart = pow(M_E, -1 * pow(input - mean, 2) / (2 * pow(stDev, 2)))
             let distribution = eulerPart / sqrt(2 * .pi) / stDev
-            
+
             return distribution
-            
+
         } else if case .multinomial = self {
-            
+
             guard let variables = variables as? [(category: Int, probability: Double)] else {
                 return nil
             }
-            
+
             guard let input = input as? Int else {
                 return nil
             }
-            
+
             return variables.first { variable in
                 return variable.category == input
                 }?.probability
-            
+
         }
-        
+
         return nil
     }
-    
+
     func train(values: [Any]) -> [Any]? {
-        
+
         if case .gaussian = self {
-            
+
             guard let values = values as? [Double] else {
                 return nil
             }
-            
+
             return [values.mean(), values.standardDeviation()]
-            
+
         } else if case .multinomial = self {
-            
+
             guard let values = values as? [Int] else {
                 return nil
             }
-            
+
             let count = values.count
             let categoryProba = values.uniques().map { value -> (Int, Double) in
                 return (value, Double(values.filter { $0 == value }.count) / Double(count))
             }
             return categoryProba
         }
-        
+
         return nil
     }
 }
 
 public class NaiveBayes<T> {
-    
+
     var variables: [Int: [(feature: Int, variables: [Any])]]
     var type: NBType
-    
+
     var data: [[T]]
     var classes: [Int]
-    
+
     public init(type: NBType, data: [[T]], classes: [Int]) throws {
         self.type = type
         self.data = data
         self.classes = classes
         self.variables = [Int: [(Int, [Any])]]()
-        
+
         if case .gaussian = type, T.self != Double.self {
             throw "When using Gaussian NB you have to have continuous features (Double)"
         } else if case .multinomial = type, T.self != Int.self {
             throw "When using Multinomial NB you have to have categorical features (Int)"
         }
     }
-    
+
     public func train() throws -> Self {
-        
+
         for `class` in classes.uniques() {
             variables[`class`] = [(Int, [Any])]()
-            
+
             let classDependent = data.enumerated().filter { (offset, _) in
                 return classes[offset] == `class`
             }
-            
+
             for feature in 0..<data[0].count {
-                
+
                 let featureDependent = classDependent.map { $0.element[feature] }
-                
+
                 guard let trained = type.train(values: featureDependent) else {
                     throw "Critical! Data could not be casted even though it was checked at init"
                 }
-                
+
                 variables[`class`]?.append((feature, trained))
             }
         }
-        
+
         return self
     }
-    
+
     public func classify(with input: [T]) -> Int {
         let likelihoods = classifyProba(with: input).max { (first, second) -> Bool in
             return first.1 < second.1
         }
-        
+
         guard let `class` = likelihoods?.0 else {
             return -1
         }
-        
+
         return `class`
     }
-    
+
     public func classifyProba(with input: [T]) -> [(Int, Double)] {
-        
+
         var probaClass = [Int: Double]()
         let amount = classes.count
-        
+
         classes.forEach { `class` in
             let individual = classes.filter { $0 == `class` }.count
             probaClass[`class`] = Double(individual) / Double(amount)
         }
-        
+
         let classesAndFeatures = variables.map { (`class`, value) -> (Int, [Double]) in
             let distribution = value.map { (feature, variables) -> Double in
                 return type.calcLikelihood(variables: variables, input: input[feature]) ?? 0.0
             }
             return (`class`, distribution)
         }
-        
+
         let likelihoods = classesAndFeatures.map { (`class`, distribution) in
             return (`class`, distribution.reduce(1, *) * (probaClass[`class`] ?? 0.0))
         }
-        
+
         let sum = likelihoods.map { $0.1 }.reduce(0, +)
         let normalized = likelihoods.map { (`class`, likelihood) in
             return (`class`, likelihood / sum)
         }
-        
+
         return normalized
     }
 }

Naive Bayes Classifier/NaiveBayes.playground/contents.xcplayground

@@ -1,4 +1,4 @@
 <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
-<playground version='5.0' target-platform='ios' display-mode='raw'>
+<playground version='5.0' target-platform='ios' display-mode='rendered'>
     <timeline fileName='timeline.xctimeline'/>
 </playground>

Naive Bayes Classifier/NaiveBayes.playground/timeline.xctimeline

@@ -17,5 +17,15 @@
          selectedRepresentationIndex = "0"
          shouldTrackSuperviewWidth = "NO">
       </LoggerValueHistoryTimelineItem>
+      <LoggerValueHistoryTimelineItem
+         documentLocation = "#CharacterRangeLen=10&amp;CharacterRangeLoc=3410&amp;EndingColumnNumber=15&amp;EndingLineNumber=100&amp;StartingColumnNumber=5&amp;StartingLineNumber=100&amp;Timestamp=514208705.303681"
+         selectedRepresentationIndex = "0"
+         shouldTrackSuperviewWidth = "NO">
+      </LoggerValueHistoryTimelineItem>
+      <LoggerValueHistoryTimelineItem
+         documentLocation = "#CharacterRangeLen=6&amp;CharacterRangeLoc=1931&amp;EndingColumnNumber=11&amp;EndingLineNumber=53&amp;StartingColumnNumber=5&amp;StartingLineNumber=53&amp;Timestamp=514208705.303681"
+         selectedRepresentationIndex = "0"
+         shouldTrackSuperviewWidth = "NO">
+      </LoggerValueHistoryTimelineItem>
    </TimelineItems>
 </Timeline>