Divepred.py

#!/usr/bin/env python
# coding: utf-8

# In[1]:


import pandas as pd
carsales=pd.read_csv('carsales.csv')


# In[2]:


print(carsales.head())


# In[1]:


carsales.columns= ['month','sales']


# In[3]:


import pandas as pd
carsales=pd.read_csv('carsales.csv')


# In[4]:


print(carsales.head())


# In[5]:


carsales.columns= ['month','sales']


# In[6]:


print(carsales.tail())


# In[7]:


carsales=carsales.loc[0:107,:].copy()


# In[8]:


carsales['period']=list(range(108))


# In[9]:


from matplotlib import pyplot as plt
plt.scatter(carsales['period'],carsales['sales'])
plt.title('Car Sales by Month')
plt.xlabel('Month')
plt.ylabel('Sales')
plt.show()


# In[10]:


x = carsales['period'].values.reshape(-1,1)
y = carsales['sales'].values.reshape(-1,1)


# In[11]:


from sklearn.linear_model import LinearRegression
regressor = LinearRegression()
regressor.fit(x, y)


# In[12]:


print(regressor.coef_)
print(regressor.intercept_)


# In[13]:


plt.scatter(carsales['period'],carsales['sales'])
plt.plot(carsales['period'],[81.2 * i + 10250.8 for i in \
carsales['period']],'r-',label='Regression Line')
plt.plot(carsales['period'],[125 * i + 8000 for i in
carsales['period']],'r--',label='Hypothesized Line')
plt.legend(loc="upper left")
plt.title('Car Sales by Month')
plt.xlabel('Month')
plt.ylabel('Sales')
plt.show()


# In[14]:


saleslist=carsales['sales'].tolist()
regressionline=[81.2 * i + 10250.8 for i in
carsales['period']]
hypothesizedline=[125 * i + 8000 for i in
carsales['period']]
error1=[(x-y) for x, y in zip(regressionline,saleslist)]
error2=[(x-y) for x, y in zip(hypothesizedline,saleslist)]


# In[15]:


print(error1)
print(error2)


# In[16]:


import numpy as np
error1abs=[abs(value) for value in error1]
error2abs=[abs(value) for value in error2]
print(np.mean(error1abs))
print(np.mean(error2abs))


# In[17]:


error1squared=[(value)**2 for value in error1]
error2squared=[(value)**2 for value in error2]
print(np.sqrt(np.mean(error1squared)))
print(np.sqrt(np.mean(error2squared)))


# In[18]:


def get_mae(line,actual):
error=[(x-y) for x,y in zip(line,actual)]
errorabs=[abs(value) for value in error]
mae=np.mean(errorabs)
return(mae)
def get_rmse(line,actual):
error=[(x-y) for x,y in zip(line,actual)]
errorsquared=[(value)**2 for value in error]
rmse=np.sqrt(np.mean(errorsquared))
return(rmse)


# In[19]:


def get_mae(line,actual):
    error=[(x-y) for x,y in zip(line,actual)]
    errorabs=[abs(value) for value in error]
    mae=np.mean(errorabs)
    return(mae)
    def get_rmse(line,actual):
    error=[(x-y) for x,y in zip(line,actual)]
    errorsquared=[(value)**2 for value in error]
    rmse=np.sqrt(np.mean(errorsquared))
    return(rmse)


# In[20]:


def get_mae(line,actual):
 error=[(x-y) for x,y in zip(line,actual)]
errorabs=[abs(value) for value in error]
mae=np.mean(errorabs)
return(mae)
def get_rmse(line,actual):
error=[(x-y) for x,y in zip(line,actual)]
errorsquared=[(value)**2 for value in error]
rmse=np.sqrt(np.mean(errorsquared))
return(rmse)


# In[21]:


x_extended = np.append(carsales['period'], np.arange(108,
116))


# In[22]:


x_extended=x_extended.reshape(-1,1)
extended_prediction=regressor.predict(x_extended)


# In[23]:


plt.scatter(carsales['period'],carsales['sales'])
plt.plot(x_extended,extended_prediction,'r--')
plt.title('Car Sales by Month')
plt.xlabel('Month')
plt.ylabel('Sales')
plt.show()


# In[24]:


carsales['quadratic']=carsales['period'].apply(lambda x:
x**2)
carsales['cubic']=carsales['period'].apply(lambda x: x**3)
x3 = carsales.loc[:,
['period','quadratic','cubic']].values.reshape(-1,3)
y = carsales['sales'].values.reshape(-1,1)
regressor_cubic = LinearRegression()
regressor_cubic.fit(x3, y)
plt.scatter(carsales['period'],carsales['sales'])
plt.plot(x,regressor.predict(x),'r-')
plt.plot(x,regressor_cubic.predict(x3),'r--')
plt.title('Car Sales by Month')
plt.xlabel('Month')
plt.ylabel('Sales')
plt.show()


# In[25]:


print(regressor_cubic.coef_)
print(regressor_cubic.intercept_)


# In[26]:


from matplotlib import pyplot as plt
plt.plot(carsales['period'],carsales['sales'])
plt.title('Car Sales by Month')
plt.xlabel('Month')
plt.ylabel('Sales')
plt.show()


# In[27]:


import math
carsales['sin_period']=carsales['period'].apply(lambda x:
math.sin(x*2*math.pi/12))
carsales['cos_period']=carsales['period'].apply(lambda x:
math.cos(x*2*math.pi/12))
x_trig = carsales.loc[:,
['period','sin_period','cos_period']].values.reshape(-1,3)
y = carsales['sales'].values.reshape(-1,1)
regressor_trig = LinearRegression()
regressor_trig.fit(x_trig, y)
plt.plot(carsales['period'],carsales['sales'])
plt.plot(x,regressor_trig.predict(x_trig),'r--')
plt.title('Car Sales by Month')
plt.xlabel('Month')
plt.ylabel('Sales')
plt.show()


# In[28]:


trig_line=regressor_trig.predict(x_trig)[:, 0]
print(get_rmse(trig_line,saleslist))


# In[29]:


carsales['squareroot']=carsales['period'].apply(lambda x:
x**0.5)
carsales['exponent15']=carsales['period'].apply(lambda x:
x**1.5)
carsales['log']=carsales['period'].apply(lambda x:
math.log(x+1))
x_complex = carsales.loc[:,
['period','log','sin_period','cos_period', \
'squareroot','exponent15','log','quadratic',
'cubic']].values.reshape(-1,9)
y = carsales['sales'].values.reshape(-1,1)
regressor_complex = LinearRegression()
regressor_complex.fit(x_complex,y)
complex_line=[prediction for sublist in
regressor_complex.predict(x_complex) \
for prediction in sublist]
print(get_rmse(complex_line,saleslist))


# In[30]:


x_complex_train = carsales.loc[0:80,
['period','log','sin_period','cos_period','squareroot', \
'exponent15','log','quadratic','cubic']].values.reshape(-1,9)
y_train = carsales.loc[0:80,'sales'].values.reshape(-1,1)
x_complex_test = carsales.loc[81:107,
['period','log','sin_period','cos_period','squareroot', \
'exponent15','log','quadratic','cubic']].values.reshape(-1,9)
y_test = carsales.loc[81:107,'sales'].values.reshape(-1,1)
regressor_complex.fit(x_complex_train, y_train)


# In[31]:


x_train = carsales.loc[0:80,
['period']].values.reshape(-1,1)
x_test = carsales.loc[81:107,
['period']].values.reshape(-1,1)
x_trig_train = carsales.loc[0:80,
['period','sin_period','cos_period']].values.reshape(-1,3)
x_trig_test = carsales.loc[81:107,
['period','sin_period','cos_period']].values.reshape(-1,3)
regressor.fit(x_train, y_train)
regressor_trig.fit(x_trig_train, y_train)
complex_test_predictions=[prediction for sublist in \
regressor_complex.predict(x_complex_test) for
prediction in sublist]
test_predictions=[prediction for sublist in
regressor.predict(x_test) for \
prediction in sublist]
trig_test_predictions=[prediction for sublist in \
regressor_trig.predict(x_trig_test) for prediction in
sublist]
print(get_rmse(test_predictions,saleslist[81:107]))
print(get_rmse(trig_test_predictions,saleslist[81:107]))
print(get_rmse(complex_test_predictions,saleslist[81:107]))


# In[ ]: