import warnings warnings.filterwarnings('ignore') #importing the libraries import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns

cars = pd.read_csv('CarPrice_Assignment.csv') cars.head()

cars.shape

cars.describe()

cars.info()

# Splitting company name from CarName column CompanyName = cars['CarName'].apply(lambda x : x.split(' ')[0]) cars.insert(3,"CompanyName",CompanyName) cars.drop(['CarName'],axis=1,inplace=True) cars.head()

cars.CompanyName.unique()

cars.CompanyName = cars.CompanyName.str.lower() def replace_name(a,b): cars.CompanyName.replace(a,b,inplace=True) replace_name('alfa-romero','alfa-romeo') replace_name('maxda','mazda') replace_name('porcshce','porsche') replace_name('toyouta','toyota') replace_name('vokswagen','volkswagen') replace_name('vw','volkswagen') cars.CompanyName.unique()

#Checking for duplicates cars.loc[cars.duplicated()]

cars.columns

plt.figure(figsize=(20,8)) plt.subplot(1,2,1) plt.title('Car Price Distribution Plot') sns.distplot(cars.price) plt.subplot(1,2,2) plt.title('Car Price Spread') sns.boxplot(y=cars.price) plt.show()

print(cars.price.describe(percentiles = [0.25,0.50,0.75,0.85,0.90,1]))

plt.figure(figsize=(25, 6)) plt.subplot(1,3,1) plt1 = cars.CompanyName.value_counts().plot(kind='bar') plt.title('Companies Histogram') plt1.set(xlabel = 'Car company', ylabel='Frequency of company') plt.subplot(1,3,2) plt1 = cars.fueltype.value_counts().plot(kind='bar') plt.title('Fuel Type Histogram') plt1.set(xlabel = 'Fuel Type', ylabel='Frequency of fuel type') plt.subplot(1,3,3) plt1 = cars.carbody.value_counts().plot(kind='bar') plt.title('Car Type Histogram') plt1.set(xlabel = 'Car Type', ylabel='Frequency of Car type') plt.show()

plt.figure(figsize=(20,8)) plt.subplot(1,2,1) plt.title('Symboling Histogram') sns.countplot(cars.symboling) plt.subplot(1,2,2) plt.title('Symboling vs Price') sns.boxplot(x=cars.symboling, y=cars.price) plt.show()

plt.figure(figsize=(20,8)) plt.subplot(1,2,1) plt.title('Engine Type Histogram') sns.countplot(cars.enginetype) plt.subplot(1,2,2) plt.title('Engine Type vs Price') sns.boxplot(x=cars.enginetype, y=cars.price)

df = pd.DataFrame(cars.groupby(['enginetype'])['price'].mean().sort_values(ascending = False)) df.plot.bar(figsize=(8,6)) plt.title('Engine Type vs Average Price') plt.show()

plt.figure(figsize=(25, 6)) df = pd.DataFrame(cars.groupby(['CompanyName'])['price'].mean().sort_values(ascending = False)) df.plot.bar() plt.title('Company Name vs Average Price') df = pd.DataFrame(cars.groupby(['fueltype'])['price'].mean().sort_values(ascending = False)) df.plot.bar() plt.title('Fuel Type vs Average Price') df = pd.DataFrame(cars.groupby(['carbody'])['price'].mean().sort_values(ascending = False)) df.plot.bar() plt.title('Car Type vs Average Price') plt.show()

plt.figure(figsize=(15,5)) plt.subplot(1,2,1) plt.title('Door Number Histogram') sns.countplot(cars.doornumber) plt.subplot(1,2,2) plt.title('Door Number vs Price') sns.boxplot(x=cars.doornumber, y=cars.price) plt.show() plt.figure(figsize=(15,5)) plt.subplot(1,2,1) plt.title('Aspiration Histogram') sns.countplot(cars.aspiration) plt.subplot(1,2,2) plt.title('Aspiration vs Price') sns.boxplot(x=cars.aspiration, y=cars.price) plt.show()

def plot_count(x,fig): plt.subplot(4,2,fig) plt.title(x+' Histogram') sns.countplot(cars[x]) plt.subplot(4,2,(fig+1)) plt.title(x+' vs Price') sns.boxplot(x=cars[x], y=cars.price) plt.figure(figsize=(15,20)) plot_count('enginelocation', 1) plot_count('cylindernumber', 3) plot_count('fuelsystem', 5) plot_count('drivewheel', 7) plt.tight_layout()

def scatter(x,fig): plt.subplot(5,2,fig) plt.scatter(cars[x],cars['price']) plt.title(x+' vs Price') plt.ylabel('Price') plt.xlabel(x) plt.figure(figsize=(10,20)) scatter('carlength', 1) scatter('carwidth', 2) scatter('carheight', 3) scatter('curbweight', 4) plt.tight_layout()

# Derive New field - Fuel economy cars['fueleconomy'] = (0.55 * cars['citympg']) + (0.45 * cars['highwaympg'])

# Binning the Car Companies based on avg prices of each Company. cars['price'] = cars['price'].astype('int') temp = cars.copy() table = temp.groupby(['CompanyName'])['price'].mean() temp = temp.merge(table.reset_index(), how='left',on='CompanyName') bins = [0,10000,20000,40000] cars_bin=['Budget','Medium','Highend'] cars['carsrange'] = pd.cut(temp['price_y'],bins,right=False,labels=cars_bin) cars.head()

plt.figure(figsize=(8,6)) plt.title('Fuel economy vs Price') sns.scatterplot(x=cars['fueleconomy'],y=cars['price'], hue=cars['drivewheel']) plt.xlabel('Fuel Economy') plt.ylabel('Price') plt.show() plt.tight_layout()

cars_lr = cars[['price', 'fueltype', 'aspiration','carbody', 'drivewheel','wheelbase', 'curbweight', 'enginetype', 'cylindernumber', 'enginesize', 'boreratio','horsepower', 'fueleconomy', 'carlength','carwidth', 'carsrange']] cars_lr.head()

sns.pairplot(cars_lr) plt.show()

# Check the corr values of final list of variables cor = cars.corr() cor

# Find out the Fields with high correlation correlated_features = set() for i in range(len(cor.columns)): for j in range(i): if abs(cor.iloc[i, j]) > 0.8: colname1 = cor.columns[i] colname2 = cor.columns[j] print(abs(cor.iloc[i, j]), "--", i, '--', j, '--', colname1, '--', colname2) correlated_features.add(colname1) correlated_features.add(colname2)

print(cor.columns) print('------') print(correlated_features)

cor['highwaympg']['citympg']

# Print the Correlation values of the High Correlated fields corh = cars[correlated_features].corr() corh

# Produce a Heatmap plt.figure(figsize=(14,14)) sns.heatmap(corh, annot=True, linewidths=.5, fmt=".2f", cmap="YlGnBu")

print(cars['fueltype'].unique()) print(cars['aspiration'].unique()) print(cars['carbody'].unique()) print(cars['drivewheel'].unique()) print(cars['enginetype'].unique()) print(cars['cylindernumber'].unique()) print(cars['carsrange'].unique()) print(cars['fuelsystem'].unique()) print(cars['CompanyName'].unique()) print(cars['doornumber'].unique()) print(cars['enginelocation'].unique())

# Defining the map function def dummies(x,df): temp = pd.get_dummies(df[x], drop_first = True) df = pd.concat([df, temp], axis = 1) df.drop([x], axis = 1, inplace = True) return df # Applying the function to the cars_lr cars_lr = cars cars_lr = dummies('fueltype',cars_lr) cars_lr = dummies('aspiration',cars_lr) cars_lr = dummies('carbody',cars_lr) cars_lr = dummies('drivewheel',cars_lr) cars_lr = dummies('enginetype',cars_lr) cars_lr = dummies('cylindernumber',cars_lr) cars_lr = dummies('carsrange',cars_lr) cars_lr = dummies('CompanyName',cars_lr) cars_lr = dummies('doornumber',cars_lr) cars_lr = dummies('enginelocation',cars_lr) cars_lr = dummies('fuelsystem',cars_lr)

cars_lr.info()

cars_lr.shape

from sklearn.model_selection import train_test_split np.random.seed(0) df_train, df_test = train_test_split(cars_lr, train_size = 0.7, test_size = 0.3, random_state = 100)

from sklearn.preprocessing import MinMaxScaler scaler = MinMaxScaler() num_vars = ['wheelbase', 'carheight', 'stroke', 'curbweight', 'enginesize', 'boreratio', 'horsepower','fueleconomy','carlength','carwidth','price'] df_train[num_vars] = scaler.fit_transform(df_train[num_vars])

df_train.head()

df_train.describe()

#Dividing data into X and y variables y_train = df_train.pop('price') X_train = df_train

!pip install statsmodels

from sklearn.feature_selection import RFE from sklearn.linear_model import LinearRegression import statsmodels.api as sm from statsmodels.stats.outliers_influence import variance_inflation_factor

def build_model(X,y): X = sm.add_constant(X) # Adding the constant lm = sm.OLS(y,X).fit() # fitting the model print(lm.summary()) # model summary return lm def checkVIF(X): vif = pd.DataFrame() vif['Features'] = X.columns vif['VIF'] = [variance_inflation_factor(X.values, i) for i in range(X.shape[1])] vif['VIF'] = round(vif['VIF'], 2) vif = vif.sort_values(by = "VIF", ascending = False) return(vif)

model1 = build_model(X_train, y_train)

X_train.columns

X_train1 = X_train.drop( ['car_ID', 'wheelbase', 'carlength', 'compressionratio', 'horsepower', 'citympg', 'highwaympg', 'hardtop', 'sedan','wagon','fwd','rwd','dohcv','l','ohc','ohcf','ohcv','six','Medium', 'audi','buick','chevrolet','dodge','honda','isuzu','jaguar','mazda','nissan','porsche','renault','saab', 'subaru','toyota','volkswagen','volvo','two','2bbl','4bbl','idi','mfi','mpfi','spdi'], axis=1)

X_train_df, model2 = build_model(X_train1, y_train)

X_train1.columns

checkVIF(X_train1)

X_train2 = X_train1.drop(['enginesize', 'peakrpm', 'curbweight', 'four', 'boreratio', 'carwidth', 'stroke', 'gas', 'fueleconomy', 'carheight', 'rotor'], axis=1)

model3 = build_model(X_train2, y_train)

lm = LinearRegression() lm.fit(X_train1,y_train) rfe = RFE(lm, 10) rfe = rfe.fit(X_train1, y_train)

list(zip(X_train.columns,rfe.support_,rfe.ranking_))

X_train1.columns[rfe.support_]

X_train_rfe = X_train[X_train1.columns[rfe.support_]] X_train_rfe.head()

model4 = build_model(X_train_rfe,y_train)

X_train_new = X_train_rfe.drop(["twelve"], axis = 1)

model5 = build_model(X_train_new,y_train)

X_train_new1 = sm.add_constant(X_train_new)

X_train_new1

y_train_price = model5.predict(X_train_new1)

# Plot the histogram of the error terms fig = plt.figure() sns.distplot((y_train - y_train_price), bins = 20) fig.suptitle('Error Terms', fontsize = 20) # Plot heading plt.xlabel('Errors', fontsize = 18)

df_test

# Scaling the test set num_vars = ['carwidth', 'curbweight', 'enginesize', 'boreratio', 'rotor', 'three', 'Highend', 'bmw', 'rear','price'] df_test1 = pd.DataFrame(scaler.fit_transform(df_test[num_vars]), columns=num_vars)

df_test1

#Dividing into X and y y_test = df_test1.pop('price') X_test = df_test1

# Now let's use our model to make predictions. X_test_new = pd.DataFrame(sm.add_constant(X_test))

X_test_new

# Making predictions y_pred = model5.predict(X_test_new)

from sklearn.metrics import r2_score print("Test Prediction R-Sqrd: ", r2_score(y_test, y_pred))

print("Test Prediction R-Sqrd: ", r2_score(y_train, y_train_price))

# Plot the histogram of the error terms fig = plt.figure() sns.distplot((y_test - y_pred), bins = 30) fig.suptitle('Error Terms', fontsize = 20) # Plot heading plt.xlabel('Errors', fontsize = 18)

#EVALUATION OF THE MODEL # Plotting y_test and y_pred to understand the spread. fig = plt.figure() plt.scatter(y_test,y_pred) fig.suptitle('y_test vs y_pred', fontsize=20) # Plot heading plt.xlabel('y_test', fontsize=18) # X-label plt.ylabel('y_pred', fontsize=16)

## Residual Normality residt = model5.resid residt probplot = sm.ProbPlot(residt) plt.figure(figsize=(8,6)) probplot.ppplot(line='45') plt.title('Normal P-P Plot for Regression Standardised Residuals') plt.show()

## Test of Homoscedasticity def get_standard_values (parm): return (parm - parm.mean())/parm.std() plt.scatter(get_standard_values(model5.fittedvalues), get_standard_values(residt)) plt.title('Residual Analysis - Noises') plt.xlabel('Standarddised Predicted Values') plt.ylabel('Standarddised Residuals')

print(model5.summary())