diction-using-logostic-regression

September 22, 2024

#Importing Libraries
[5]: import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')

[6]: import seaborn as sns

#Load Dataset
[7]: df = pd.read_csv('german_credit_data.csv')

The selected attributes are:

Age (numeric) Sex (text: male, female) Job (numeric: 0 - unskilled and non-resident, 1 - unskilled
and resident, 2 - skilled, 3 - highly skilled) Housing (text: own, rent, or free) Saving accounts (text
- little, moderate, quite rich, rich) Checking account (numeric, in DM - Deutsch Mark) Credit
amount (numeric, in DM) Duration (numeric, in month) Purpose (text: car, furniture/equipment,
radio/TV, domestic appliances, repairs, education, business, vacation/others)
#Data Preparation
[8]: df.head(10)

[8]: Unnamed: 0 Age Sex Job Housing Saving accounts Checking account \
0 0 67 male 2 own NaN little
1 1 22 female 2 own little moderate
2 2 49 male 1 own little NaN
3 3 45 male 2 free little little
4 4 53 male 2 free little little
5 5 35 male 1 free NaN NaN
6 6 53 male 2 own quite rich NaN
7 7 35 male 3 rent little moderate
8 8 61 male 1 own rich NaN
9 9 28 male 3 own little moderate

Credit amount Duration Purpose

1
 0 1169 6 radio/TV
1 5951 48 radio/TV
2 2096 12 education
3 7882 42 furniture/equipment
4 4870 24 car
5 9055 36 education
6 2835 24 furniture/equipment
7 6948 36 car
8 3059 12 radio/TV
9 5234 30 car

[9]: df.shape

[9]: (1000, 10)

[10]: df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 10 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Unnamed: 0 1000 non-null int64
1 Age 1000 non-null int64
2 Sex 1000 non-null object
3 Job 1000 non-null int64
4 Housing 1000 non-null object
5 Saving accounts 817 non-null object
6 Checking account 606 non-null object
7 Credit amount 1000 non-null int64
8 Duration 1000 non-null int64
9 Purpose 1000 non-null object
dtypes: int64(5), object(5)
memory usage: 78.2+ KB

[11]: df.describe(include='all')

[11]: Unnamed: 0 Age Sex Job Housing Saving accounts \

count 1000.000000 1000.000000 1000 1000.000000 1000 817
unique NaN NaN 2 NaN 3 4
top NaN NaN male NaN own little
freq NaN NaN 690 NaN 713 603
mean 499.500000 35.546000 NaN 1.904000 NaN NaN
std 288.819436 11.375469 NaN 0.653614 NaN NaN
min 0.000000 19.000000 NaN 0.000000 NaN NaN
25% 249.750000 27.000000 NaN 2.000000 NaN NaN
50% 499.500000 33.000000 NaN 2.000000 NaN NaN

2
 75% 749.250000 42.000000 NaN 2.000000 NaN NaN
max 999.000000 75.000000 NaN 3.000000 NaN NaN

Checking account Credit amount Duration Purpose

count 606 1000.000000 1000.000000 1000
unique 3 NaN NaN 8
top little NaN NaN car
freq 274 NaN NaN 337
mean NaN 3271.258000 20.903000 NaN
std NaN 2822.736876 12.058814 NaN
min NaN 250.000000 4.000000 NaN
25% NaN 1365.500000 12.000000 NaN
50% NaN 2319.500000 18.000000 NaN
75% NaN 3972.250000 24.000000 NaN
max NaN 18424.000000 72.000000 NaN

#EDA
[12]: df.hist(figsize=(20, 20))
plt.show()

3
[13]: df = df.loc[:, ~df.columns.str.contains('^Unnamed')]

[14]: df.to_csv('german_credit_data.csv')

[15]: w=df['Credit amount'].max()

print(w)

18424

[16]: def determine_creditability(row):

# Example criteria (adjust these based on your requirements)
if row['Credit amount'] < 5000: # Replace with your threshold
return 1 # Creditworthy

4
 else:
return 0 # Not creditworthy

# Apply the function to each row

df['Creditability'] = df.apply(determine_creditability, axis=1)

# Save the updated DataFrame

df.to_csv('updated_file.csv', index=False)

[17]: df1=pd.read_csv('updated_file.csv')
df.head(2)

[17]: Age Sex Job Housing Saving accounts Checking account Credit amount \
0 67 male 2 own NaN little 1169
1 22 female 2 own little moderate 5951

Duration Purpose Creditability

0 6 radio/TV 1
1 48 radio/TV 0

Here 1-credit worthy 2-not credit worthy

[18]: sns.set_style('darkgrid')
plt.ylabel("total no of credit defaulters")
dx=sns.countplot(x='Creditability', data=df1)
plt.show()

5
[19]: plt.figure(figsize=(20,10))
sns.countplot(x='Duration', hue='Creditability', data=df1)
plt.xlabel('Duration')
plt.ylabel('total no of credit defaulters')
plt.show()

6
[20]: gender_df=df1.groupby(["Sex", "Creditability"])["Purpose"].value_counts()
gender_df

[20]: Sex Creditability Purpose

female 0 car 21
radio/TV 8
furniture/equipment 7
business 4
education 3
vacation/others 2
1 radio/TV 77
car 73
furniture/equipment 67
education 21
business 15
domestic appliances 6
repairs 5
vacation/others 1
male 0 car 68
business 22
radio/TV 19
furniture/equipment 15
education 9
vacation/others 6
repairs 4
1 radio/TV 176
car 175
furniture/equipment 92

7
 business 56
education 26
repairs 13
domestic appliances 6
vacation/others 3
Name: count, dtype: int64

[21]: plt.figure(figsize=(20,10))
ax=sns.countplot(x='Sex', hue='Job', data=df1)
plt.legend(title='Job',loc='upper right')
plt.show()

[22]: creditability_counts = df1.groupby(['Sex', 'Creditability']).size().

↪unstack(fill_value=0)

colors = ['#FF9999', '#66B3FF']

# Plot pie charts for each gender
for gender in creditability_counts.index:
plt.figure(figsize=(8, 8))
plt.pie(creditability_counts.loc[gender], labels=['Not Creditworthy (0)',␣
↪'Creditworthy (1)'], autopct='%1.1f%%', colors=colors)

plt.title(f'Creditability Distribution for {gender.capitalize()}')

plt.show()

8
9
[23]: plt.figure(figsize=(30,10))
plt.ylabel("total no of credit defaulters")
dx=sns.countplot(x='Purpose', hue='Creditability', data=df1)
plt.show()

10
 1: Good customers 2: Bad customers
[24]: plt.figure(figsize=(15,8))
plt.ylabel("total no of credit defaulters")
dx=sns.countplot(x='Creditability', hue='Housing', data=df1)
plt.show()

Correlation
[25]: corr = df1[['Creditability', 'Credit amount']].corr()

# Create a heatmap
plt.figure(figsize=(8, 6))
sns.heatmap(corr, annot=True, cmap='coolwarm', fmt='.2f', square=True)
plt.title('Correlation Heatmap')
plt.show()

11
 converting categorical value into numerical values
[26]: df1['Sex'] = df1['Sex'].replace({'male': 1, 'female': 0})

[27]: df1['Housing'] = df1['Housing'].replace({'own': 0, 'free': 1, 'rent': 2})

[28]: options=['little' 'moderate', 'rich', 'quite rich']

nan_mask = df1['Saving accounts'].isna()
df1.loc[nan_mask, 'Saving accounts'] = np.random.choice(options, size=nan_mask.
↪sum())

[29]: nan_mask = df1['Checking account'].isna()

df1.loc[nan_mask, 'Checking account'] = np.random.choice(options, size=nan_mask.
↪sum())

[30]: df1['Saving accounts'] = df1['Saving accounts'].replace({'little': 0,␣

↪'moderate': 1, 'rich': 2, 'quite rich': 3})

12
 df1['Checking account'] = df1['Checking account'].replace({'little': 0,␣
↪'moderate': 1, 'rich': 2, 'quite rich': 3})

[31]: df1['Saving accounts']=df1['Saving accounts'].replace({'littlemoderate':4})

df1['Checking account']=df1['Checking account'].replace({'littlemoderate':4})

[32]: df1.drop(['Purpose'],axis=1,inplace=True)

#preparation of datasets
[33]: Predictor = df1[df1.columns[df1.columns != 'Creditability']] # All columns␣
↪except 'Creditability'

Target = df1['Creditability']

Splitting data for training and testing

[34]: from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(Predictor, Target,␣
↪test_size = 0.3, random_state = 1)

[35]: print("X train",X_train.shape)

print("Y train",y_train.shape)
print("X test",X_test.shape)
print("y test",y_test.shape)

X train (700, 8)
Y train (700,)
X test (300, 8)
y test (300,)
#Model Training
[36]: from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, confusion_matrix,␣
↪classification_report

[37]: logr=LogisticRegression()
logr.fit(X_train,y_train)
y_Pred=logr.predict(X_test)

[38]: y_Pred

[38]: array([0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1,
1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0,
1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1,
1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1,

13
 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1,
1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, 1,
1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1,
1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1,
1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1])

[39]: conf_matrix=confusion_matrix(y_test,y_Pred)
conf_matrix

[39]: array([[ 56, 6],

[ 8, 230]])

**Confusion matrix structure

[40]: #([[True positive(TP), False Negative(FN)],
# [False Positive(FP), True Negavtive(TN)]

#True Positives (TP): This value should be high. It indicates the number of␣
↪actual positives that were correctly identified by the model.

#True Negatives (TN): This value should also be high. It indicates the number␣
↪of actual negatives that were correctly identified.

#False Positives (FP): This value should be low. It indicates the number of␣
↪actual negatives that were incorrectly classified as positives.

#False Negatives (FN): This value should be low. It indicates the number of␣
↪actual positives that were incorrectly classified as negatives.

[41]: sns.heatmap(pd.DataFrame(conf_matrix), annot=True, cmap="YlGnBu" ,fmt='g')

plt.title('Confusion matrix')
plt.ylabel('Actual label')
plt.xlabel('Predicted label')

[41]: Text(0.5, 23.52222222222222, 'Predicted label')

14
 Accuracy = (TP+TN)/(TP+TN+FP+FN)

Precision = (TP)/(TP+FP)

Recall = (TP)/(TP+FN)

F1 score = 2 x (Precision x Recall)/(Precision + Recall)

[42]: print('Accuracy',accuracy_score(y_test,y_Pred))
print('Precision',logr.score(X_test,y_test))
print('Recall',logr.score(X_test,y_test))
print('F1 score',logr.score(X_test,y_test))

Accuracy 0.9533333333333334
Precision 0.9533333333333334
Recall 0.9533333333333334
F1 score 0.9533333333333334

15
[43]: print(classification_report(y_test,y_Pred))

precision recall f1-score support

0 0.88 0.90 0.89 62

1 0.97 0.97 0.97 238

accuracy 0.95 300

macro avg 0.92 0.93 0.93 300
weighted avg 0.95 0.95 0.95 300

[46]: from sklearn import metrics

[48]: y_pred_proba=logr.predict_proba(X_test)[:,1]
fpr, tpr, _ = metrics.roc_curve(y_test, y_pred_proba)
auc = metrics.roc_auc_score(y_test, y_pred_proba)

[51]: plt.plot(fpr,tpr,label="data 1, auc="+str(auc))

plt.legend(loc=4)
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("ROC Curve")
plt.show()

16
[53]: from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score

[55]: kfold=KFold(n_splits=3, random_state=7, shuffle=True)

result = cross_val_score(logr, X_train, y_train, cv=kfold, scoring='accuracy')
print(result.mean())

0.9528447232309892
The Model has accuracy rate of 95%
[58]: df1.to_csv('preprocessed_data.csv', index=False)

17