import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import seaborn as sns

df1 = pd.read_csv('Training_Data_Set.csv')
df1.head(3)

Age Vroom
Owner
ID Maker model Location Distance manufacture_year of engine_displacement engine_power body_type Audit transmi
Type
car Rating

0 11100001 skoda octavia Ahmedabad NaN Second 1964.0 55 1964 147.0 compact 8

1 11100002 fiat panda Ahmedabad 27750.0 Third 2012.0 7 1242 51.0 NaN 6

2 11100003 bmw x1 Hyderabad 46000.0 Third 2014.0 5 1995 105.0 NaN 7

df = df1.drop(['ID'] , axis = 1)
df.head(3)

Age Vroom
Owner
Maker model Location Distance manufacture_year of engine_displacement engine_power body_type Audit transmission
Type
car Rating

0 skoda octavia Ahmedabad NaN Second 1964.0 55 1964 147.0 compact 8 man

1 fiat panda Ahmedabad 27750.0 Third 2012.0 7 1242 51.0 NaN 6 man

2 bmw x1 Hyderabad 46000.0 Third 2014.0 5 1995 105.0 NaN 7 auto

(df.describe())

Distance manufacture_year Age of car engine_displacement engine_power Vroom Audit Rating Price

count 5.230400e+04 53513.000000 53515.000000 53515.000000 52076.000000 53515.000000 5.351500e+04

mean 9.454626e+04 2010.408032 8.591890 1904.049014 100.448345 5.998374 1.098084e+06

std 2.755617e+05 4.650367 4.650322 1496.564596 45.330622 1.418336 8.441565e+05

min 0.000000e+00 1934.000000 3.000000 14.000000 10.000000 4.000000 3.000000e+00

25% 1.549000e+04 2008.000000 5.000000 1395.000000 73.000000 5.000000 5.051812e+05

50% 6.552000e+04 2011.000000 8.000000 1896.000000 91.000000 6.000000 8.854552e+05

75% 1.356410e+05 2014.000000 11.000000 1995.000000 125.000000 7.000000 1.477829e+06

max 9.899800e+06 2016.000000 85.000000 32000.000000 896.000000 8.000000 2.212078e+07

df.shape

(53515, 16)

for column in df.columns:

if df[column].dtype =='object':
print(column.upper(),':',df[column].nunique())
print(df[column].value_counts().sort_values())
print('\n')

MAKER : 8
maserati 38
fiat 1845
hyundai 2240
nissan 5485
bmw 7178
audi 7326
toyota 7840
skoda 21563
Name: Maker, dtype: int64

MODEL : 23
tt 903
juke 955
citigo 1120
q7 1245
roomster 1322
rapid 1409
aygo 1486
avensis 1512
auris 1666
micra 1676
coupe 1710
q3 1736
panda 1769
yeti 1898
x5 1979
q5 2039
i30 2047
x1 2420
x3 2779
qashqai 2854
yaris 3176
superb 3195
octavia 12619
Name: model, dtype: int64

LOCATION : 11
Ahmedabad 4770
Hyderabad 4804
Delhi 4822
Chennai 4834
Mumbai 4860
Pune 4862
Kolkata 4867
Jaipur 4870
Bangalore 4877
Kochi 4969
Coimbatore 4974
Name: Location, dtype: int64

OWNER TYPE : 4
Fourth & Above 13349
Second 13365
Third 13395
First 13406
Name: Owner Type, dtype: int64

BODY_TYPE : 2
van 9
compact 4127
Name: body_type, dtype: int64

TRANSMISSION : 2
auto 16781
man 36734
Name: transmission, dtype: int64

DOOR_COUNT : 7
1 2
6 8
3 185
2 4348
None 7534
5 7630
4 33808
Name: door_count, dtype: int64

SEAT_COUNT : 10
1 1
8 1
9 2
6 23
3 109
2 725
7 852
4 4467
None 8511
5 38824
Name: seat_count, dtype: int64

FUEL_TYPE : 2
petrol 25956
diesel 27559
Name: fuel_type, dtype: int64

df = pd.get_dummies(df , columns = ['Maker','model','Location','Owner Type','body_type','transmission','door_count'

df
 Age Vroom
Distance manufacture_year of engine_displacement engine_power Audit Price Maker_bmw Maker_fiat Maker_hyundai ... seat
car Rating

0 NaN 1964.0 55 1964 147.0 8 543764.25 0 0 0 ...

1 27750.0 2012.0 7 1242 51.0 6 401819.25 0 1 0 ...

2 46000.0 2014.0 5 1995 105.0 7 2392855.50 1 0 0 ...

3 43949.0 2011.0 8 1618 140.0 7 958606.50 0 0 0 ...

4 59524.0 2012.0 7 2993 180.0 7 3085561.50 1 0 0 ...

... ... ... ... ... ... ... ... ... ... ... ...

53510 29334.0 2014.0 5 1598 77.0 4 1342996.50 0 0 0 ...

53511 223631.0 2009.0 10 1900 77.0 8 510732.75 0 0 0 ...

53512 25500.0 2015.0 4 1995 105.0 4 2008123.50 1 0 0 ...

53513 1195500.0 2011.0 8 11950 93.0 5 874352.25 0 0 0 ...

53514 142000.0 2008.0 11 2993 173.0 4 1576610.25 1 0 0 ...

53515 rows × 67 columns

df.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 53515 entries, 0 to 53514
Data columns (total 67 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Distance 52304 non-null float64
1 manufacture_year 53513 non-null float64
2 Age of car 53515 non-null int64
3 engine_displacement 53515 non-null int64
4 engine_power 52076 non-null float64
5 Vroom Audit Rating 53515 non-null int64
6 Price 53515 non-null float64
7 Maker_bmw 53515 non-null uint8
8 Maker_fiat 53515 non-null uint8
9 Maker_hyundai 53515 non-null uint8
10 Maker_maserati 53515 non-null uint8
11 Maker_nissan 53515 non-null uint8
12 Maker_skoda 53515 non-null uint8
13 Maker_toyota 53515 non-null uint8
14 model_avensis 53515 non-null uint8
15 model_aygo 53515 non-null uint8
16 model_citigo 53515 non-null uint8
17 model_coupe 53515 non-null uint8
18 model_i30 53515 non-null uint8
19 model_juke 53515 non-null uint8
20 model_micra 53515 non-null uint8
21 model_octavia 53515 non-null uint8
22 model_panda 53515 non-null uint8
23 model_q3 53515 non-null uint8
24 model_q5 53515 non-null uint8
25 model_q7 53515 non-null uint8
26 model_qashqai 53515 non-null uint8
27 model_rapid 53515 non-null uint8
28 model_roomster 53515 non-null uint8
29 model_superb 53515 non-null uint8
30 model_tt 53515 non-null uint8
31 model_x1 53515 non-null uint8
32 model_x3 53515 non-null uint8
33 model_x5 53515 non-null uint8
34 model_yaris 53515 non-null uint8
35 model_yeti 53515 non-null uint8
36 Location_Bangalore 53515 non-null uint8
37 Location_Chennai 53515 non-null uint8
38 Location_Coimbatore 53515 non-null uint8
39 Location_Delhi 53515 non-null uint8
40 Location_Hyderabad 53515 non-null uint8
41 Location_Jaipur 53515 non-null uint8
42 Location_Kochi 53515 non-null uint8
43 Location_Kolkata 53515 non-null uint8
44 Location_Mumbai 53515 non-null uint8
45 Location_Pune 53515 non-null uint8
46 Owner Type_Fourth & Above 53515 non-null uint8
47 Owner Type_Second 53515 non-null uint8
48 Owner Type_Third 53515 non-null uint8
49 body_type_van 53515 non-null uint8
50 transmission_man 53515 non-null uint8
51 door_count_2 53515 non-null uint8
52 door_count_3 53515 non-null uint8
53 door_count_4 53515 non-null uint8
54 door_count_5 53515 non-null uint8
55 door_count_6 53515 non-null uint8
56 door_count_None 53515 non-null uint8
57 seat_count_2 53515 non-null uint8
58 seat_count_3 53515 non-null uint8
59 seat_count_4 53515 non-null uint8
60 seat_count_5 53515 non-null uint8
61 seat_count_6 53515 non-null uint8
62 seat_count_7 53515 non-null uint8
63 seat_count_8 53515 non-null uint8
64 seat_count_9 53515 non-null uint8
65 seat_count_None 53515 non-null uint8
66 fuel_type_petrol 53515 non-null uint8
dtypes: float64(4), int64(3), uint8(60)
memory usage: 5.9 MB

df.isnull().sum()

Distance 1211
manufacture_year 2
Age of car 0
engine_displacement 0
engine_power 1439
...
seat_count_7 0
seat_count_8 0
seat_count_9 0
seat_count_None 0
fuel_type_petrol 0
Length: 67, dtype: int64
df = df.fillna(df.mean())

df.isnull().sum()

Distance 0
manufacture_year 0
Age of car 0
engine_displacement 0
engine_power 0
..
seat_count_7 0
seat_count_8 0
seat_count_9 0
seat_count_None 0
fuel_type_petrol 0
Length: 67, dtype: int64

df.duplicated().sum()

114

df1=df.drop_duplicates()
df1.head(2)

Age Vroom
Distance manufacture_year of engine_displacement engine_power Audit Price Maker_bmw Maker_fiat Maker_hyundai ... seat_c
car Rating

0 94546.262446 1964.0 55 1964 147.0 8 543764.25 0 0 0 ...

1 27750.000000 2012.0 7 1242 51.0 6 401819.25 0 1 0 ...

2 rows × 67 columns

sns.heatmap(df1.iloc[:, 0:6].corr(),annot=True)
plt.show()

import matplotlib.pyplot as plt

import seaborn as sns

# Assuming df1 is your DataFrame

feature_list = df1.columns

# Calculate the number of rows and columns needed

num_plots = len(feature_list)
rows = (num_plots // 5) + (1 if num_plots % 5 else 0)
cols = 5

plt.figure(figsize=(20, rows * 4))

for i in range(num_plots):
plt.subplot(rows, cols, i + 1)
sns.boxplot(y=df1[feature_list[i]], data=df1)
plt.title('Boxplot of {}'.format(feature_list[i]))

plt.tight_layout()
plt.show()
def remove_outlier(col):
Q1,Q3=col.quantile([0.25,0.75])
IQR=Q3-Q1
lower_range= Q1-(1.5*IQR)
upper_range= Q3+(1.5*IQR)
return lower_range,upper_range

for i in feature_list:
LL, UL = remove_outlier(df1[i])
df1[i] = np.where(df1[i] > UL,UL, df1[i])
df1[i] = np.where(df1[i] < LL,LL,df1[i])

/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:3: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead
See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret
urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] < LL,LL,df1[i])
/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:3: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret

urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] > UL,UL, df1[i])
/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:4: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret

urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] < LL,LL,df1[i])
/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:3: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret

urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] > UL,UL, df1[i])
/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:4: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret

urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] < LL,LL,df1[i])
/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:3: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret

urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] > UL,UL, df1[i])
/var/folders/y7/v00504rn5c5dgvzgtlt4bkcr0000gn/T/ipykernel_3490/610677943.py:4: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#ret

urning-a-view-versus-a-copy
df1[i] = np.where(df1[i] < LL,LL,df1[i])

import matplotlib.pyplot as plt

import seaborn as sns

# Assuming df1 is your DataFrame

feature_list = df1.columns

# Calculate the number of rows and columns needed

num_plots = len(feature_list)
rows = (num_plots // 5) + (1 if num_plots % 5 else 0)
cols = 5

plt.figure(figsize=(20, rows * 4))

for i in range(num_plots):
plt.subplot(rows, cols, i + 1)
sns.boxplot(y=df1[feature_list[i]], data=df1)
plt.title('Boxplot of {}'.format(feature_list[i]))

plt.tight_layout()
plt.show()
Pair plot
#sns.pairplot(df1 , diag_kind = 'kde')

Train Test split

# Copy all the predictor variable into X dataframe

X = df1.drop('Price', axis=1)

# Copy target into the y dataframe.

y = df1[['Price']]

# Split X and y into training and test set in 75:25 ratio

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X,y, test_size=0.25, random_state=1)

import statsmodels.api as sm

X_train=sm.add_constant(X_train)
X_test=sm.add_constant(X_test)

model = sm.OLS(y_train,X_train).fit()
model.summary()

OLS Regression Results

Dep. Variable: Price R-squared: 0.790

Model: OLS Adj. R-squared: 0.790

Method: Least Squares F-statistic: 1.255e+04

Date: Wed, 05 Jun 2024 Prob (F-statistic): 0.00

Time: 08:04:04 Log-Likelihood: -5.6646e+05

No. Observations: 40050 AIC: 1.133e+06

Df Residuals: 40037 BIC: 1.133e+06

Df Model: 12

Covariance Type: nonrobust

 coef std err t P>|t| [0.025 0.975]

const -5.881e+07 1.47e+08 -0.400 0.690 -3.47e+08 2.3e+08

Distance -3.2001 0.036 -89.621 0.000 -3.270 -3.130

manufacture_year 2.962e+04 7.29e+04 0.406 0.685 -1.13e+05 1.73e+05

Age of car -2.463e+04 7.29e+04 -0.338 0.735 -1.68e+05 1.18e+05

engine_displacement 254.1679 6.738 37.723 0.000 240.962 267.374

engine_power 6632.8889 78.444 84.556 0.000 6479.138 6786.640

Vroom Audit Rating 963.0607 1182.810 0.814 0.416 -1355.275 3281.396

Maker_bmw 3.57e-08 2.9e-09 12.308 0.000 3e-08 4.14e-08

Maker_fiat 3.856e-09 3.1e-10 12.438 0.000 3.25e-09 4.46e-09

Maker_hyundai -1.203e-08 9.75e-10 -12.343 0.000 -1.39e-08 -1.01e-08

Maker_maserati -2.032e-10 1.77e-11 -11.452 0.000 -2.38e-10 -1.68e-10

Maker_nissan -2.264e-11 2.05e-12 -11.021 0.000 -2.67e-11 -1.86e-11

Maker_skoda -1.416e+05 3740.872 -37.854 0.000 -1.49e+05 -1.34e+05

Maker_toyota 0 0 nan nan 0 0

model_avensis 0 0 nan nan 0 0

model_aygo 0 0 nan nan 0 0

model_citigo 0 0 nan nan 0 0

model_coupe 0 0 nan nan 0 0

model_i30 0 0 nan nan 0 0

model_juke 0 0 nan nan 0 0

model_micra 0 0 nan nan 0 0

model_octavia 0 0 nan nan 0 0

model_panda 0 0 nan nan 0 0

model_q3 0 0 nan nan 0 0

model_q5 0 0 nan nan 0 0

model_q7 0 0 nan nan 0 0

model_qashqai 0 0 nan nan 0 0

model_rapid 0 0 nan nan 0 0

model_roomster 0 0 nan nan 0 0

model_superb 0 0 nan nan 0 0

model_tt 0 0 nan nan 0 0

model_x1 0 0 nan nan 0 0

model_x3 0 0 nan nan 0 0

model_x5 0 0 nan nan 0 0

model_yaris 0 0 nan nan 0 0

model_yeti 0 0 nan nan 0 0

Location_Bangalore 0 0 nan nan 0 0

Location_Chennai 0 0 nan nan 0 0

 Location_Coimbatore 0 0 nan nan 0 0

Location_Delhi 0 0 nan nan 0 0

Location_Hyderabad 0 0 nan nan 0 0

Location_Jaipur 0 0 nan nan 0 0

Location_Kochi 0 0 nan nan 0 0

Location_Kolkata 0 0 nan nan 0 0

Location_Mumbai 0 0 nan nan 0 0

Location_Pune 0 0 nan nan 0 0

Owner Type_Fourth & Above 0 0 nan nan 0 0

Owner Type_Second 0 0 nan nan 0 0

Owner Type_Third -243.6636 3864.482 -0.063 0.950 -7818.138 7330.811

body_type_van 0 0 nan nan 0 0

transmission_man -2.246e+05 4440.331 -50.585 0.000 -2.33e+05 -2.16e+05

door_count_2 0 0 nan nan 0 0

door_count_3 0 0 nan nan 0 0

door_count_4 -8.997e+04 4011.925 -22.427 0.000 -9.78e+04 -8.21e+04

door_count_5 0 0 nan nan 0 0

door_count_6 0 0 nan nan 0 0

door_count_None 0 0 nan nan 0 0

seat_count_2 0 0 nan nan 0 0

seat_count_3 0 0 nan nan 0 0

seat_count_4 0 0 nan nan 0 0

seat_count_5 6.139e+04 4130.664 14.863 0.000 5.33e+04 6.95e+04

seat_count_6 0 0 nan nan 0 0

seat_count_7 0 0 nan nan 0 0

seat_count_8 0 0 nan nan 0 0

seat_count_9 0 0 nan nan 0 0

seat_count_None 0 0 nan nan 0 0

fuel_type_petrol -1.971e+05 4201.344 -46.921 0.000 -2.05e+05 -1.89e+05

Omnibus: 6020.661 Durbin-Watson: 2.009

Prob(Omnibus): 0.000 Jarque-Bera (JB): 16656.071

Skew: 0.818 Prob(JB): 0.00

Kurtosis: 5.702 Cond. No. 1.36e+16

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The smallest eigenvalue is 2.76e-18. This might indicate that there are
strong multicollinearity problems or that the design matrix is singular.

# Lets check the VIP(variation inflation factor)of the predictors

from statsmodels.stats.outliers_influence import variance_inflation_factor

vif_series1 = pd.Series(
 [variance_inflation_factor(X_train.values, i) for i in range(X_train.shape[1])],
index=X_train.columns,
)
vif_file = "VIF values: \n\n{}\n".format(vif_series1)
print(vif_file)

/Users/sachinssharma/anaconda3/lib/python3.11/site-packages/statsmodels/regression/linear_model.py:1781: Runtim
eWarning: invalid value encountered in scalar divide
return 1 - self.ssr/self.centered_tss
VIF values:

const 7.682852e+09
Distance 2.604749e+00
manufacture_year 3.463081e+04
Age of car 3.463332e+04
engine_displacement 4.206017e+00
...
seat_count_7 NaN
seat_count_8 NaN
seat_count_9 NaN
seat_count_None NaN
fuel_type_petrol 1.563505e+00
Length: 67, dtype: float64

my_df = vif_series1.to_frame(name='col_names')
my_df.to_excel('vif.xlsx')

my_df.head(10)

col_names

const 7.682852e+09

Distance 2.604749e+00

manufacture_year 3.463081e+04

Age of car 3.463332e+04

engine_displacement 4.206017e+00

engine_power 3.461206e+00

Vroom Audit Rating 1.000134e+00

Maker_bmw NaN

Maker_fiat NaN

Maker_hyundai NaN

#1) Removing predictor 'engine_displacement' as VIF>2

X_train2=X_train.drop(["engine_displacement"], axis=1)
olsmod_1=sm.OLS(y_train, X_train2)
olsres_1=olsmod_1.fit()
print(
"R-squared:",
np.round(olsres_1.rsquared, 3),
"\nAdjusted R-squared:",
np.round(olsres_1.rsquared_adj, 3),
)

R-squared: 0.783
Adjusted R-squared: 0.782

0.790-0.783

0.007000000000000006

#1) Removing predictor 'Age of car' as VIF>2

X_train2=X_train.drop(["Age of car"], axis=1)
olsmod_1=sm.OLS(y_train, X_train2)
olsres_1=olsmod_1.fit()
print(
"R-squared:",
np.round(olsres_1.rsquared, 3),
"\nAdjusted R-squared:",
np.round(olsres_1.rsquared_adj, 3),
)

R-squared: 0.79
Adjusted R-squared: 0.79

0.790-0.79

0.0

#1) Removing predictor 'manufacture_year' as VIF>2

X_train2=X_train.drop(["manufacture_year"], axis=1)
olsmod_1=sm.OLS(y_train, X_train2)
olsres_1=olsmod_1.fit()
print(
"R-squared:",
np.round(olsres_1.rsquared, 3),
"\nAdjusted R-squared:",
np.round(olsres_1.rsquared_adj, 3),
)
R-squared: 0.79
Adjusted R-squared: 0.79

#1) Removing predictor 'Distance ' as VIF>2

X_train2=X_train.drop(["Distance"], axis=1)
olsmod_1=sm.OLS(y_train, X_train2)
olsres_1=olsmod_1.fit()
print(
"R-squared:",
np.round(olsres_1.rsquared, 3),
"\nAdjusted R-squared:",
np.round(olsres_1.rsquared_adj, 3),
)

R-squared: 0.748
Adjusted R-squared: 0.748

0.790-0.748 #Don't remove

0.04200000000000004

#1) Removing predictor 'engine_power' as VIF>2

X_train2=X_train.drop(["engine_power"], axis=1)
olsmod_1=sm.OLS(y_train, X_train2)
olsres_1=olsmod_1.fit()
print(
"R-squared:",
np.round(olsres_1.rsquared, 3),
"\nAdjusted R-squared:",
np.round(olsres_1.rsquared_adj, 3),
)

R-squared: 0.753
Adjusted R-squared: 0.752

0.790-0.753 # don't remove

0.03700000000000003

#As we are observing the multicollinearity and no such diffrence in removing manufacture_year , Age of car , engine_d
# so remove and run regression model

Dropping Multicolinear columns

X_train = X_train.drop(["manufacture_year"], axis = 1)

olsmod_5 = sm.OLS(y_train,X_train)
olsres_5 = olsmod_5.fit()
olsres_5.summary()

OLS Regression Results

Dep. Variable: Price R-squared: 0.790

Model: OLS Adj. R-squared: 0.790

Method: Least Squares F-statistic: 1.369e+04

Date: Wed, 05 Jun 2024 Prob (F-statistic): 0.00

Time: 08:04:09 Log-Likelihood: -5.6646e+05

No. Observations: 40050 AIC: 1.133e+06

Df Residuals: 40038 BIC: 1.133e+06

Df Model: 11

Covariance Type: nonrobust

coef std err t P>|t| [0.025 0.975]

const 1e+06 1.37e+04 72.895 0.000 9.73e+05 1.03e+06

Distance -3.2001 0.036 -89.621 0.000 -3.270 -3.130

Age of car -5.425e+04 621.554 -87.289 0.000 -5.55e+04 -5.3e+04

engine_displacement 254.1727 6.738 37.724 0.000 240.967 267.379

engine_power 6632.8456 78.443 84.557 0.000 6479.096 6786.595

Vroom Audit Rating 964.6015 1182.792 0.816 0.415 -1353.698 3282.901

 Maker_bmw -2.141e-09 1.39e-09 -1.546 0.122 -4.86e-09 5.74e-10

Maker_fiat 1.325e-09 8.61e-10 1.539 0.124 -3.63e-10 3.01e-09

Maker_hyundai -9.028e-10 6.21e-10 -1.454 0.146 -2.12e-09 3.14e-10

Maker_maserati -3.339e-13 2.45e-13 -1.363 0.173 -8.14e-13 1.46e-13

Maker_nissan -4.951e-13 5.14e-12 -0.096 0.923 -1.06e-11 9.59e-12

Maker_skoda -1.416e+05 3740.810 -37.853 0.000 -1.49e+05 -1.34e+05

Maker_toyota 0 0 nan nan 0 0

model_avensis 0 0 nan nan 0 0

model_aygo 0 0 nan nan 0 0

model_citigo 0 0 nan nan 0 0

model_coupe 0 0 nan nan 0 0

model_i30 0 0 nan nan 0 0

model_juke 0 0 nan nan 0 0

model_micra 0 0 nan nan 0 0

model_octavia 0 0 nan nan 0 0

model_panda 0 0 nan nan 0 0

model_q3 0 0 nan nan 0 0

model_q5 0 0 nan nan 0 0

model_q7 0 0 nan nan 0 0

model_qashqai 0 0 nan nan 0 0

model_rapid 0 0 nan nan 0 0

model_roomster 0 0 nan nan 0 0

model_superb 0 0 nan nan 0 0

model_tt 0 0 nan nan 0 0

model_x1 0 0 nan nan 0 0

model_x3 0 0 nan nan 0 0

model_x5 0 0 nan nan 0 0

model_yaris 0 0 nan nan 0 0

model_yeti 0 0 nan nan 0 0

Location_Bangalore 0 0 nan nan 0 0

Location_Chennai 0 0 nan nan 0 0

Location_Coimbatore 0 0 nan nan 0 0

Location_Delhi 0 0 nan nan 0 0

Location_Hyderabad 0 0 nan nan 0 0

Location_Jaipur 0 0 nan nan 0 0

Location_Kochi 0 0 nan nan 0 0

Location_Kolkata 0 0 nan nan 0 0

Location_Mumbai 0 0 nan nan 0 0

Location_Pune 0 0 nan nan 0 0

Owner Type_Fourth & Above 0 0 nan nan 0 0

Owner Type_Second 0 0 nan nan 0 0

Owner Type_Third -239.5045 3864.428 -0.062 0.951 -7813.873 7334.864

body_type_van 0 0 nan nan 0 0

transmission_man -2.246e+05 4440.243 -50.587 0.000 -2.33e+05 -2.16e+05

door_count_2 0 0 nan nan 0 0

door_count_3 0 0 nan nan 0 0

door_count_4 -8.998e+04 4011.878 -22.428 0.000 -9.78e+04 -8.21e+04

door_count_5 0 0 nan nan 0 0

door_count_6 0 0 nan nan 0 0

door_count_None 0 0 nan nan 0 0

seat_count_2 0 0 nan nan 0 0

seat_count_3 0 0 nan nan 0 0

seat_count_4 0 0 nan nan 0 0

seat_count_5 6.139e+04 4130.620 14.863 0.000 5.33e+04 6.95e+04

seat_count_6 0 0 nan nan 0 0

 seat_count_7 0 0 nan nan 0 0

seat_count_8 0 0 nan nan 0 0

seat_count_9 0 0 nan nan 0 0

seat_count_None 0 0 nan nan 0 0

fuel_type_petrol -1.971e+05 4201.233 -46.920 0.000 -2.05e+05 -1.89e+05

Omnibus: 6020.713 Durbin-Watson: 2.009

Prob(Omnibus): 0.000 Jarque-Bera (JB): 16655.937

Skew: 0.818 Prob(JB): 0.00

Kurtosis: 5.702 Cond. No. 1.36e+16

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The smallest eigenvalue is 2.76e-18. This might indicate that there are
strong multicollinearity problems or that the design matrix is singular.

#by removing engine_displacement and age of car we are getting 0.783 and 0.745 respectively so remove only manufactur

#So Remove only manufacturer year

#columns = Vroom Audit Rating,Maker_bmw,Maker_fiat,Maker_hyundai,Maker_maserati,Maker_nissan,Owner Type_Third are hav

x_train5 = X_train.drop(['Vroom Audit Rating','Maker_bmw','Maker_fiat','Maker_hyundai','Maker_maserati','Maker_nissan

olsmod_6 = sm.OLS(y_train,x_train5)
olsres_6 = olsmod_6.fit()
olsres_6.summary()

OLS Regression Results

Dep. Variable: Price R-squared: 0.790

Model: OLS Adj. R-squared: 0.790

Method: Least Squares F-statistic: 1.674e+04

Date: Wed, 05 Jun 2024 Prob (F-statistic): 0.00

Time: 08:04:10 Log-Likelihood: -5.6646e+05

No. Observations: 40050 AIC: 1.133e+06

Df Residuals: 40040 BIC: 1.133e+06

Df Model: 9

Covariance Type: nonrobust

coef std err t P>|t| [0.025 0.975]

const 1.006e+06 1.17e+04 85.729 0.000 9.83e+05 1.03e+06

Distance -3.2001 0.036 -89.626 0.000 -3.270 -3.130

Age of car -5.425e+04 621.532 -87.292 0.000 -5.55e+04 -5.3e+04

engine_displacement 254.1769 6.737 37.727 0.000 240.972 267.382

engine_power 6632.8515 78.441 84.559 0.000 6479.106 6786.598

Maker_skoda -1.416e+05 3740.706 -37.854 0.000 -1.49e+05 -1.34e+05

Maker_toyota -6.021e-10 4.01e-11 -15.023 0.000 -6.81e-10 -5.24e-10

model_avensis -4.284e-10 3.03e-11 -14.123 0.000 -4.88e-10 -3.69e-10

model_aygo 6.391e-13 2.39e-13 2.677 0.007 1.71e-13 1.11e-12

model_citigo 0 0 nan nan 0 0

model_coupe 0 0 nan nan 0 0

model_i30 0 0 nan nan 0 0

model_juke 0 0 nan nan 0 0

model_micra 0 0 nan nan 0 0

model_octavia 0 0 nan nan 0 0

model_panda 0 0 nan nan 0 0

model_q3 0 0 nan nan 0 0

model_q5 0 0 nan nan 0 0

model_q7 0 0 nan nan 0 0

model_qashqai 0 0 nan nan 0 0

 model_rapid 0 0 nan nan 0 0

model_roomster 0 0 nan nan 0 0

model_superb 0 0 nan nan 0 0

model_tt 0 0 nan nan 0 0

model_x1 0 0 nan nan 0 0

model_x3 0 0 nan nan 0 0

model_x5 0 0 nan nan 0 0

model_yaris 0 0 nan nan 0 0

model_yeti 0 0 nan nan 0 0

Location_Bangalore 0 0 nan nan 0 0

Location_Chennai 0 0 nan nan 0 0

Location_Coimbatore 0 0 nan nan 0 0

Location_Delhi 0 0 nan nan 0 0

Location_Hyderabad 0 0 nan nan 0 0

Location_Jaipur 0 0 nan nan 0 0

Location_Kochi 0 0 nan nan 0 0

Location_Kolkata 0 0 nan nan 0 0

Location_Mumbai 0 0 nan nan 0 0

Location_Pune 0 0 nan nan 0 0

Owner Type_Fourth & Above 0 0 nan nan 0 0

Owner Type_Second 0 0 nan nan 0 0

body_type_van 0 0 nan nan 0 0

transmission_man -2.246e+05 4440.076 -50.585 0.000 -2.33e+05 -2.16e+05

door_count_2 0 0 nan nan 0 0

door_count_3 0 0 nan nan 0 0

door_count_4 -8.999e+04 4011.542 -22.434 0.000 -9.79e+04 -8.21e+04

door_count_5 0 0 nan nan 0 0

door_count_6 0 0 nan nan 0 0

door_count_None 0 0 nan nan 0 0

seat_count_2 0 0 nan nan 0 0

seat_count_3 0 0 nan nan 0 0

seat_count_4 0 0 nan nan 0 0

seat_count_5 6.14e+04 4130.527 14.866 0.000 5.33e+04 6.95e+04

seat_count_6 0 0 nan nan 0 0

seat_count_7 0 0 nan nan 0 0

seat_count_8 0 0 nan nan 0 0

seat_count_9 0 0 nan nan 0 0

seat_count_None 0 0 nan nan 0 0

fuel_type_petrol -1.971e+05 4201.104 -46.925 0.000 -2.05e+05 -1.89e+05

Omnibus: 6022.274 Durbin-Watson: 2.009

Prob(Omnibus): 0.000 Jarque-Bera (JB): 16663.303

Skew: 0.819 Prob(JB): 0.00

Kurtosis: 5.703 Cond. No. 1.36e+16

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The smallest eigenvalue is 2.76e-18. This might indicate that there are
strong multicollinearity problems or that the design matrix is singular.

After dropping the features causes multicolinearity and the statistical insignificant ones, Our
model performance hasn't dropped sharply. this shows that these varibales did not have much
predictive power.

For linear Regression we need to check if the following assumptions hold

 1. Linearity
2. independence
3. Homoscedasticity
4. Normality of error terms
5. No strong multicolinearity

df_pred = pd.DataFrame()

df_pred["Actual values"] = y_train.values.flatten() #actual values

df_pred["Fitted values"] = olsres_6.fittedvalues.values # predicted values
df_pred["Residuals"] = olsres_6.resid.values # residuals (actual-fitted)

df_pred.head()

Actual values Fitted values Residuals

0 817741.500 7.875369e+05 30204.635729

1 2938661.625 2.520180e+06 418481.249276

2 2893229.250 2.136881e+06 756347.908289

3 74262.000 -2.064326e+04 94905.255319

4 1727234.250 1.767253e+06 -40018.473918

# let us plot the fitted values vs residuals

sns.set_style("whitegrid")
sns.residplot(
data = df_pred , x = "Fitted values" , y = "Residuals" , color = "purple" , lowess = True
)
plt.xlabel("FITTED VALUES")
plt.ylabel("Residuals")
plt.title("FITTED V/S RESIDUAL PLOT")
plt.show()

# it is having no pattern that means it is linearity in nature . assumed that linearity and independence of predictor

Test for normality

from scipy import stats
stats.shapiro(df_pred["Residuals"])
#null hypothesis : it is normally distributed

/Users/sachinssharma/anaconda3/lib/python3.11/site-packages/scipy/stats/_morestats.py:1816: UserWarning: p-valu

e may not be accurate for N > 5000.
warnings.warn("p-value may not be accurate for N > 5000.")
ShapiroResult(statistic=0.9637130498886108, pvalue=0.0)

score is less than 0.05 so reject null hypothesis, it is not normal as per the shapiro's test

Test for Homoscedasticity

import statsmodels.stats.api as sms
sms.het_goldfeldquandt(df_pred['Residuals'] , x_train5)[1]

0.7863209998227594

since p value is more than 0.05 we can say residuals are homoscedastic

#
# dropping columns from the test data that are not there in the above analysis, since by dropping both columns,

x_test2 = X_test.drop(['manufacture_year','Vroom Audit Rating','Maker_bmw','Maker_fiat','Maker_hyundai','Maker_masera

x_test2.head(2)

Age
const Distance of engine_displacement engine_power Maker_skoda Maker_toyota model_avensis model_aygo model_citigo ... seat_c
car

40755 1.0 155000.0 14.0 1995.0 110.0 0.0 0.0 0.0 0.0 0.0 ...

53337 1.0 199963.0 7.0 1968.0 81.0 1.0 0.0 0.0 0.0 0.0 ...

2 rows × 59 columns

# let's make the predictions on the test set

y_pred_test = olsres_6.predict(x_test2)
y_pred_train = olsres_6.predict(x_train5)

# to check model performance

from sklearn.metrics import mean_absolute_error , mean_squared_error

#let's check the RMSE on the train data

rmse1 = np.sqrt(mean_squared_error(y_train,y_pred_train))
rmse1

336013.3899990611

#let's check the RMSE on the test data

rmse2 = np.sqrt(mean_squared_error(y_test , y_pred_test))
rmse2

329590.62845699233

more or less both values are similar so the model is accurate