Car Sales

car.csv

CarSales.py

'''
All possible library which is going to be used
here, in this small project.
'''
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
#download the file
cars=pd.read_csv('cars.csv')
# check first five records from the Top 
print(cars.head())
# check last five records from the bottom
print(cars.tail())
# information of dataset
print(cars.info())
# drop column if it is not important,if you want
# 
drop_columns=['Acceleration','Weight']
cars=cars.drop(drop_columns,axis=1)
print(cars.info())
#drop duplicate rows if any
cars=cars.drop_duplicates(keep='first')
print(cars)
print(cars.shape)
#statistical information
print(cars.describe())
#sorting variables according to 'MPG'
cars_sort=cars.sort_values(by='MPG',ascending=False).head()
print(cars_sort)

# replace space from column
cars.columns=cars.columns.str.replace(' ','')
def min_max_value(col):
    top=cars[col].idxmax()
    top_obs=pd.DataFrame(cars.loc[top])
    
    bot=cars[col].idxmin()
    bot_obs=pd.DataFrame(cars.loc[bot])
    top_bot_obs=pd.concat([top_obs,bot_obs],axis=1)
    return top_bot_obs
max_val=min_max_value('MPG')
print(max_val)

#histogram of continuous variable
num_bins=10
plt.hist(cars['MPG'],bins=num_bins)

sns.distplot(cars['MPG'],bins=num_bins)

plt.hist(cars['Horsepower'],bins=num_bins)

# categorical data
origin_data=cars.groupby('Origin').size()
print(origin_data)
origin_data.plot(title='Origin Data')
#collect all numerical variables
cars_num=cars.select_dtypes(include=['float64','int64'])
print(cars_num.head())
cars_num.hist(bins=20)

#correlation with the variable of interest
cars_corr=cars_num.corr()['MPG'][:-1]
print(cars_corr)
#correlation plot using pair plot
for i in range(0,len(cars_num.columns),5):
     sns.pairplot(data=cars_num,y_vars=['MPG'],x_vars=cars_num.columns[i:i+5])
    
# Plotting significant correlation in one plot (heatmap)
corr=cars_num.drop('MPG',axis=1).corr()
print(corr)
sns.heatmap(corr[(corr>=0.5)|(corr<=0.4)],cmap='viridis',vmax=1.0,vmin=1.0,linewidths=1.0,annot=True)
box1=sns.boxplot(x='Cylinders',y='MPG',data=cars)
#regression plot
sns.regplot(cars['MPG'],cars['Cylinders'])