Exploratotry Data Analysis of Car Price Prediction

Tayyab Rajput

Life Cycle of a Machine Learning Project:

• 1: Understanding the Problem Statement
• 2: Data Collection
• 3: EDA (Exploratory Data Analysis)
• 4: Data Cleaning
• 5: Data Pre-Processing
• 6: Model Training
• 7: Choose Best Model

A) Problem Statement for Prediction

• The dataset comprises used cars sold on cardekho.com in India as well as important features of these cars.
• If user can predict the price of the car based on input features.
• Prediction results can be used to give new seller the price suggestion based on market condition.

Importing data and required packages

import pandas as pd 
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import warnings 
warnings.filterwarnings("ignore")

Load the dataset

df=pd.read_csv("Desktop/Datasets/cardekho_dataset.csv")
df.head()

	Unnamed: 0	car_name	brand	model	vehicle_age	km_driven	seller_type	fuel_type	transmission_type	mileage	engine	max_power	seats	selling_price
0	0	Maruti Alto	Maruti	Alto	9	120000	Individual	Petrol	Manual	19.70	796	46.30	5	120000
1	1	Hyundai Grand	Hyundai	Grand	5	20000	Individual	Petrol	Manual	18.90	1197	82.00	5	550000
2	2	Hyundai i20	Hyundai	i20	11	60000	Individual	Petrol	Manual	17.00	1197	80.00	5	215000
3	3	Maruti Alto	Maruti	Alto	9	37000	Individual	Petrol	Manual	20.92	998	67.10	5	226000
4	4	Ford Ecosport	Ford	Ecosport	6	30000	Dealer	Diesel	Manual	22.77	1498	98.59	5	570000

show top five rows of the dataset

df.head()

	Unnamed: 0	car_name	brand	model	vehicle_age	km_driven	seller_type	fuel_type	transmission_type	mileage	engine	max_power	seats	selling_price
0	0	Maruti Alto	Maruti	Alto	9	120000	Individual	Petrol	Manual	19.70	796	46.30	5	120000
1	1	Hyundai Grand	Hyundai	Grand	5	20000	Individual	Petrol	Manual	18.90	1197	82.00	5	550000
2	2	Hyundai i20	Hyundai	i20	11	60000	Individual	Petrol	Manual	17.00	1197	80.00	5	215000
3	3	Maruti Alto	Maruti	Alto	9	37000	Individual	Petrol	Manual	20.92	998	67.10	5	226000
4	4	Ford Ecosport	Ford	Ecosport	6	30000	Dealer	Diesel	Manual	22.77	1498	98.59	5	570000

Delete Columns Unnamed:0 and again show Top 5 Rows

df.drop(columns=["Unnamed: 0"],axis=1,inplace=True)

df.head()

	car_name	brand	model	vehicle_age	km_driven	seller_type	fuel_type	transmission_type	mileage	engine	max_power	seats	selling_price
0	Maruti Alto	Maruti	Alto	9	120000	Individual	Petrol	Manual	19.70	796	46.30	5	120000
1	Hyundai Grand	Hyundai	Grand	5	20000	Individual	Petrol	Manual	18.90	1197	82.00	5	550000
2	Hyundai i20	Hyundai	i20	11	60000	Individual	Petrol	Manual	17.00	1197	80.00	5	215000
3	Maruti Alto	Maruti	Alto	9	37000	Individual	Petrol	Manual	20.92	998	67.10	5	226000
4	Ford Ecosport	Ford	Ecosport	6	30000	Dealer	Diesel	Manual	22.77	1498	98.59	5	570000

Show Bottom 5 Rows

df.tail()

	car_name	brand	model	vehicle_age	km_driven	seller_type	fuel_type	transmission_type	mileage	engine	max_power	seats	selling_price
15406	Hyundai i10	Hyundai	i10	9	10723	Dealer	Petrol	Manual	19.81	1086	68.05	5	250000
15407	Maruti Ertiga	Maruti	Ertiga	2	18000	Dealer	Petrol	Manual	17.50	1373	91.10	7	925000
15408	Skoda Rapid	Skoda	Rapid	6	67000	Dealer	Diesel	Manual	21.14	1498	103.52	5	425000
15409	Mahindra XUV500	Mahindra	XUV500	5	3800000	Dealer	Diesel	Manual	16.00	2179	140.00	7	1225000
15410	Honda City	Honda	City	2	13000	Dealer	Petrol	Automatic	18.00	1497	117.60	5	1200000

Shape of the Dataset

df.shape

(15411, 13)

Show features/columns of the Dataset

df.columns

Index(['car_name', 'brand', 'model', 'vehicle_age', 'km_driven', 'seller_type',
       'fuel_type', 'transmission_type', 'mileage', 'engine', 'max_power',
       'seats', 'selling_price'],
      dtype='object')

# 2nd method 
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 15411 entries, 0 to 15410
Data columns (total 13 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   car_name           15411 non-null  object 
 1   brand              15411 non-null  object 
 2   model              15411 non-null  object 
 3   vehicle_age        15411 non-null  int64  
 4   km_driven          15411 non-null  int64  
 5   seller_type        15411 non-null  object 
 6   fuel_type          15411 non-null  object 
 7   transmission_type  15411 non-null  object 
 8   mileage            15411 non-null  float64
 9   engine             15411 non-null  int64  
 10  max_power          15411 non-null  float64
 11  seats              15411 non-null  int64  
 12  selling_price      15411 non-null  int64  
dtypes: float64(2), int64(5), object(6)
memory usage: 1.5+ MB

Summary of Data

df.describe()

	vehicle_age	km_driven	mileage	engine	max_power	seats	selling_price
count	15411.000000	1.541100e+04	15411.000000	15411.000000	15411.000000	15411.000000	1.541100e+04
mean	6.036338	5.561648e+04	19.701151	1486.057751	100.588254	5.325482	7.749711e+05
std	3.013291	5.161855e+04	4.171265	521.106696	42.972979	0.807628	8.941284e+05
min	0.000000	1.000000e+02	4.000000	793.000000	38.400000	0.000000	4.000000e+04
25%	4.000000	3.000000e+04	17.000000	1197.000000	74.000000	5.000000	3.850000e+05
50%	6.000000	5.000000e+04	19.670000	1248.000000	88.500000	5.000000	5.560000e+05
75%	8.000000	7.000000e+04	22.700000	1582.000000	117.300000	5.000000	8.250000e+05
max	29.000000	3.800000e+06	33.540000	6592.000000	626.000000	9.000000	3.950000e+07

Check Null Value Counts and DataTypes of the features

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 15411 entries, 0 to 15410
Data columns (total 13 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   car_name           15411 non-null  object 
 1   brand              15411 non-null  object 
 2   model              15411 non-null  object 
 3   vehicle_age        15411 non-null  int64  
 4   km_driven          15411 non-null  int64  
 5   seller_type        15411 non-null  object 
 6   fuel_type          15411 non-null  object 
 7   transmission_type  15411 non-null  object 
 8   mileage            15411 non-null  float64
 9   engine             15411 non-null  int64  
 10  max_power          15411 non-null  float64
 11  seats              15411 non-null  int64  
 12  selling_price      15411 non-null  int64  
dtypes: float64(2), int64(5), object(6)
memory usage: 1.5+ MB

Check duplicate values

df.duplicated().sum()

167

Check the number of unique values of each column

df.nunique()

car_name              121
brand                  32
model                 120
vehicle_age            24
km_driven            3688
seller_type             3
fuel_type               5
transmission_type       2
mileage               411
engine                110
max_power             342
seats                   8
selling_price        1086
dtype: int64

C) Exploring Data

Show Categories in columns

for i in df[['seller_type','fuel_type', 'transmission_type','seats']]:
    print(f"The catagories in '{i}' are : ",list(df[i].unique()))

The catagories in 'seller_type' are :  ['Individual', 'Dealer', 'Trustmark Dealer']
The catagories in 'fuel_type' are :  ['Petrol', 'Diesel', 'CNG', 'LPG', 'Electric']
The catagories in 'transmission_type' are :  ['Manual', 'Automatic']
The catagories in 'seats' are :  [5, 8, 7, 6, 4, 2, 9, 0]

Check Cars with 0 Seats

df[df.seats == 0]

	car_name	brand	model	vehicle_age	km_driven	seller_type	fuel_type	transmission_type	mileage	engine	max_power	seats	selling_price
3217	Honda City	Honda	City	18	40000	Individual	Petrol	Manual	13.00	1493	100.00	0	115000
12619	Nissan Kicks	Nissan	Kicks	2	10000	Individual	Diesel	Manual	19.39	1461	108.49	0	1154000

Define Numerical and Categorical Columns

cat_columns=df.select_dtypes(include="O").columns.to_list()
cat_columns

['car_name', 'brand', 'model', 'seller_type', 'fuel_type', 'transmission_type']

num_columns=df.select_dtypes(include=np.number).columns.to_list()
num_columns

['vehicle_age',
 'km_driven',
 'mileage',
 'engine',
 'max_power',
 'seats',
 'selling_price']

proportion of count data on categorical columns

for i in cat_columns:
    print("Column is ",i)
    print(df[i].value_counts(normalize=True))
    print("\n\n")

Column is  car_name
Hyundai i20           0.058789
Maruti Swift Dzire    0.057751
Maruti Swift          0.050678
Maruti Alto           0.050483
Honda City            0.049121
                        ...   
Mercedes-AMG C        0.000065
Tata Altroz           0.000065
Ferrari GTC4Lusso     0.000065
Hyundai Aura          0.000065
Force Gurkha          0.000065
Name: car_name, Length: 121, dtype: float64



Column is  brand
Maruti           0.323924
Hyundai          0.193498
Honda            0.096360
Mahindra         0.065602
Toyota           0.051457
Ford             0.051262
Volkswagen       0.040231
Renault          0.034780
BMW              0.028486
Tata             0.027902
Mercedes-Benz    0.021867
Skoda            0.021673
Audi             0.012459
Datsun           0.011031
Jaguar           0.003828
Land Rover       0.003309
Jeep             0.002660
Kia              0.002076
Porsche          0.001363
Volvo            0.001298
MG               0.001233
Mini             0.001103
Nissan           0.000714
Lexus            0.000649
Isuzu            0.000519
Bentley          0.000195
Maserati         0.000130
ISUZU            0.000130
Ferrari          0.000065
Mercedes-AMG     0.000065
Rolls-Royce      0.000065
Force            0.000065
Name: brand, dtype: float64



Column is  model
i20            0.058789
Swift Dzire    0.057751
Swift          0.050678
Alto           0.050483
City           0.049121
                 ...   
Ghibli         0.000065
Altroz         0.000065
GTC4Lusso      0.000065
Aura           0.000065
Gurkha         0.000065
Name: model, Length: 120, dtype: float64



Column is  seller_type
Dealer              0.618973
Individual          0.369801
Trustmark Dealer    0.011226
Name: seller_type, dtype: float64



Column is  fuel_type
Petrol      0.495944
Diesel      0.481409
CNG         0.019532
LPG         0.002855
Electric    0.000260
Name: fuel_type, dtype: float64



Column is  transmission_type
Manual       0.793265
Automatic    0.206735
Name: transmission_type, dtype: float64

D) Univariate Analysis

# Histogram

plt.figure(figsize=(15,15))
plt.suptitle("Univariate Analysis of Numerical Features",fontsize=20,fontweight='bold',alpha=0.8,y=1.)
for i in range(len(num_columns)):
    plt.subplot(3,3,i+1)
    sns.kdeplot(df[num_columns[i]],shade=True,cbar=True)
plt.show()

sns.boxplot(x="seats",y="selling_price",data=df)
sns.set(rc={"figure.figsize":(15,10)})
plt.show()

cat1 = ['brand','seller_type','fuel_type','transmission_type']
plt.figure(figsize=(20,15))
plt.suptitle("Univariate Analysis of Categorical Features",fontsize=20,fontweight='bold',alpha=0.8,y=1.)
for i in range(len(cat1)):
    plt.subplot(2,2,i+1)
    sns.countplot(df[cat1[i]])
    plt.xticks(rotation=55)
plt.show()

E) Multivariate Analysis

Multivariate Analysis is the analysis of more than one variable.

Check Multicollinearity of Numerical Features

df[num_columns].corr()

	vehicle_age	km_driven	mileage	engine	max_power	seats	selling_price
vehicle_age	1.000000	0.333891	-0.257394	0.098965	0.005208	0.030791	-0.241851
km_driven	0.333891	1.000000	-0.105239	0.192885	0.044421	0.192830	-0.080030
mileage	-0.257394	-0.105239	1.000000	-0.632987	-0.533128	-0.440280	-0.305549
engine	0.098965	0.192885	-0.632987	1.000000	0.807368	0.551236	0.585844
max_power	0.005208	0.044421	-0.533128	0.807368	1.000000	0.172257	0.750236
seats	0.030791	0.192830	-0.440280	0.551236	0.172257	1.000000	0.115033
selling_price	-0.241851	-0.080030	-0.305549	0.585844	0.750236	0.115033	1.000000

sns.heatmap(df[num_columns].corr(),annot=True)
plt.show()

F) Visualization

F.1) Visualize the Target Feature

Selling Price Distribution

sns.histplot(df.selling_price)

<AxesSubplot:xlabel='selling_price', ylabel='Count'>

B.2) Most Selling Cars

df.car_name.value_counts().head(10)

Hyundai i20           906
Maruti Swift Dzire    890
Maruti Swift          781
Maruti Alto           778
Honda City            757
Maruti Wagon R        717
Hyundai Grand         580
Toyota Innova         545
Hyundai Verna         492
Hyundai i10           410
Name: car_name, dtype: int64

sns.barplot(x=df.car_name.value_counts().head(10).index,y=df.car_name.value_counts().head(10).values,data=df)
plt.xlabel("Car_Name",fontsize=15)
plt.ylabel("Saled_Cars",fontsize=15)
plt.title("Top 10 most sold cars",fontsize=15)
plt.xticks(rotation= 45)
plt.show()

Check mean price of Hyundai i20 which is most sold

Hyundai_mean=df[df.car_name == "Hyundai i20"].selling_price.mean()
print("The mean price of Hyundai i20 is :",round(Hyundai_mean,2),"rupees")

The mean price of Hyundai i20 is : 543603.75 rupees

B.3) Most Selling Brands

df.brand.value_counts().head(10)

Maruti        4992
Hyundai       2982
Honda         1485
Mahindra      1011
Toyota         793
Ford           790
Volkswagen     620
Renault        536
BMW            439
Tata           430
Name: brand, dtype: int64

sns.barplot(x=df.brand.value_counts().head(10).index,y=df.brand.value_counts().head(10).values)
plt.xticks(rotation = 45)
plt.show()

Check the Mean price of Maruti brand which is most sold

maruti_mean=df[df.brand == "Maruti"].selling_price.mean()
print("The mean of the Maruti brand is : ",round(maruti_mean,2),"Rupees")

The mean of the Maruti brand is :  487089.32 Rupees

B.4) Costliest Brand and Costliest Car

df.groupby(["brand"])["selling_price"].max().sort_values().reset_index()

	brand	selling_price
0	Datsun	650000
1	Force	700000
2	Renault	1155000
3	Maruti	1225000
4	Volkswagen	1250000
5	Nissan	1450000
6	Tata	1750000
7	ISUZU	1900000
8	MG	2075000
9	Isuzu	2300000
10	Hyundai	2600000
11	Mahindra	2950000
12	Ford	3200000
13	Honda	3200000
14	Kia	3525000
15	Skoda	3550000
16	Toyota	3650000
17	Mini	3875000
18	Mercedes-AMG	5100000
19	Jeep	5600000
20	Maserati	6200000
21	Jaguar	6300000
22	Audi	6800000
23	Lexus	8000000
24	Volvo	8195000
25	BMW	8500000
26	Land Rover	9200000
27	Porsche	11100000
28	Mercedes-Benz	13000000
29	Bentley	14500000
30	Rolls-Royce	24200000
31	Ferrari	39500000

costly_barnd=df.groupby(["brand"])["selling_price"].max().reset_index()
sns.barplot(x=costly_barnd.brand,y=costly_barnd.selling_price)
plt.xticks(rotation = 45)
plt.show()

B.5) Costliest Car

pd.pivot_table(df,index="car_name",values=["selling_price"],aggfunc="max").sort_values(by="selling_price",ascending=False).head(10)

	selling_price
car_name
Ferrari GTC4Lusso	39500000
Rolls-Royce Ghost	24200000
Bentley Continental	14500000
Mercedes-Benz S-Class	13000000
Porsche Cayenne	11100000
Land Rover Rover	9200000
BMW 7	8500000
BMW Z4	8250000
Volvo XC	8195000
BMW X5	8100000

costly_cars=pd.pivot_table(df,index="car_name",values=["selling_price"],aggfunc="max").sort_values(by="selling_price",ascending=False).head(10).reset_index()
sns.barplot(x=costly_cars["car_name"],y=costly_cars["selling_price"])
plt.xticks(rotation = 90)
plt.show()

Most Mileage Brand and Car Name

pd.pivot_table(df , index="brand", values="mileage",aggfunc="max").sort_values(by="mileage",ascending=False).head(10)

	mileage
brand
Maruti	33.54
Hyundai	30.48
Honda	27.40
Tata	27.28
Ford	26.10
Mahindra	25.32
Renault	25.17
Toyota	23.87
Mini	23.80
Datsun	23.00

brand_mileage=pd.pivot_table(df , index="brand", values="mileage",aggfunc="max").sort_values(by="mileage",ascending=False).head(10).reset_index()
sns.barplot(x=brand_mileage.brand , y =brand_mileage.mileage)
plt.xticks(rotation = 45)
plt.title("Brand vs Mileage",fontsize=15)
plt.xlabel("Brand Name",fontsize=15)
plt.ylabel("Mileage in kmpl",fontsize=15)
plt.show()

Car with Highest Mileage

pd.pivot_table(df , index="car_name", values="mileage",aggfunc="max").sort_values(by="mileage",ascending=False).head(10)

	mileage
car_name
Maruti Wagon R	33.54
Maruti Alto	33.44
Maruti Celerio	31.79
Hyundai Santro	30.48
Maruti Swift Dzire	28.40
Maruti Swift	28.40
Maruti Ciaz	28.09
Honda Amaze	27.40
Maruti Baleno	27.39
Honda Jazz	27.30

car_mileage=pd.pivot_table(df , index="car_name", values="mileage",aggfunc="max").sort_values(by="mileage",ascending=False).head(10).reset_index()
sns.barplot(car_mileage.car_name , y= car_mileage.mileage)
plt.xticks(rotation = 45)
plt.title("Car vs Mileage",fontsize=15)
plt.xlabel("Car Name",fontsize=15)
plt.ylabel("Mileage in kmpl",fontsize=15)
plt.show()

Kilometer driven vs Selling Price

sns.scatterplot(x=df.km_driven , y = df.selling_price , hue ="fuel_type" , data = df)
plt.show()

pd.pivot_table(df,index="fuel_type",values="selling_price",aggfunc="max")

	selling_price
fuel_type
CNG	1080000
Diesel	9200000
Electric	2000000
LPG	420000
Petrol	39500000

sns.barplot(x=df.fuel_type , y = df.selling_price)
plt.title("Fuel type vs Selling_price",fontsize=20)
plt.show()

Most sold Fuel type

df.fuel_type.value_counts()

Petrol      7643
Diesel      7419
CNG          301
LPG           44
Electric       4
Name: fuel_type, dtype: int64

sns.barplot(x=df.fuel_type.value_counts().index , y = df.fuel_type.value_counts().values)
plt.title("Car vs Mileage")
plt.xlabel("Car Name")
plt.ylabel("Mileage in kmpl")
plt.show()

Fuel types available and mileage given

# mean mileage
pd.pivot_table(df , index="fuel_type", values="mileage",aggfunc=np.mean).sort_values(by="mileage",ascending=False)

	mileage
fuel_type
CNG	25.814651
Diesel	20.060030
Electric	19.160000
Petrol	19.123045
LPG	17.836364

sns.boxplot(x=df.fuel_type , y = df.mileage)
plt.title("Fuel vs Mileage",fontsize= 15)
plt.xlabel("Fuel Type",fontsize=15)
plt.ylabel("Mileage in kmpl",fontsize=15)
plt.show()

Mileage vs Selling Price

sns.scatterplot(x=df.mileage , y = df.selling_price , hue="fuel_type" , data=df)
plt.title("Mileage vs Selling_Price",fontsize= 15)
plt.xlabel("Mileage",fontsize=15)
plt.ylabel("Selling_price",fontsize=15)
plt.show()

Mileage Distribution

sns.histplot(df.mileage,kde=True,color="g")
plt.title("Mileage Distribution",fontsize= 15)
plt.xlabel("Mileage",fontsize=15)
plt.ylabel("Count",fontsize=15)
plt.show()

Vechile age vs Selling_price

sns.lineplot(x=df.vehicle_age , y=df.selling_price)
plt.title("Vehicle_age vs Selling_price",fontsize= 15)
plt.xlabel("Vehicle_age",fontsize=15)
plt.ylabel("Selling_price",fontsize=15)
plt.show()

Vehicle Age vs Mileage

result=pd.pivot_table(df , index="vehicle_age", values="mileage",aggfunc=np.max).sort_values(by="mileage",ascending=True)
result.query("vehicle_age==[29,0,3,4,5]")

	mileage
vehicle_age
29	22.05
0	22.48
5	33.54
4	33.54
3	33.54

sns.boxplot(x= df.vehicle_age , y = df.mileage)
plt.title("Vehicle_age vs Mileage",fontsize= 15)
plt.xlabel("Vehicle_age",fontsize=15)
plt.ylabel("Mileage",fontsize=15)
plt.show()

Car_name vs Age

pd.pivot_table(df , index="car_name", values="vehicle_age",aggfunc="max").sort_values(by="vehicle_age",ascending=False).head(10)

	vehicle_age
car_name
Maruti Alto	29
BMW 3	25
Honda City	22
Maruti Wagon R	21
Mahindra Bolero	18
Mahindra Scorpio	18
Skoda Octavia	18
Honda CR-V	17
Mercedes-Benz E-Class	17
Honda Civic	15

Transmission Type :

sns.countplot(df.transmission_type)
plt.title("Transmission Type Count",fontsize= 15)
plt.xlabel("Transmission_Type",fontsize=15)
plt.ylabel("Count",fontsize=15)
plt.show()

Transmission type vs Selling Price

pd.pivot_table(df , index="transmission_type", values="selling_price",aggfunc="max").sort_values(by="selling_price",ascending=False).head(10)

	selling_price
transmission_type
Automatic	39500000
Manual	4050000

sns.barplot(x=df.transmission_type , y = df.selling_price , data = df)
plt.title("Transmission Type vs Price",fontsize= 15)
plt.xlabel("Transmission_Type",fontsize=15)
plt.ylabel("Selling_Price in Millions",fontsize=15)
plt.show()

Seller Type

pd.pivot_table(df , index="seller_type", values="selling_price",aggfunc="median").astype("float")

	selling_price
seller_type
Dealer	591000.0
Individual	507000.0
Trustmark Dealer	540000.0

sns.barplot(x=df.seller_type , y = df.selling_price , data = df)
plt.title("Seller Type vs Price",fontsize= 15)
plt.xlabel("Seller_Type",fontsize=15)
plt.ylabel("Selling_Price in Millions",fontsize=15)
plt.show()