In [1]: %matplotlib inline
Major Project: Credit Risk Modeling –
Loan Classification
Name: Divyasnh Date: 2025-06-22
This project aims to classify loan applicants based on their likelihood of repaying
loans using various machine learning models.
In [3]: import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
from sklearn.model_selection import train_test_split, StratifiedKFold, cross_val_score
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.impute import SimpleImputer
from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
import xgboost as xgb
from sklearn.metrics import classification_report, confusion_matrix, roc_auc_score
import shap
warnings.filterwarnings("ignore")
sns.set(style='whitegrid')
In [5]: # Load the dataset
df = pd.read_csv('loan_detection.csv')
print("Dataset shape:", df.shape)
df.head()
Dataset shape: (41188, 60)
Out[5]:
age campaign pdays previous no_previous_contact not_working job_admin.
0 56 1 999 0 1 0 0
1 57 1 999 0 1 0 0
2 37 1 999 0 1 0 0
3 40 1 999 0 1 0 1
4 56 1 999 0 1 0 0
5 rows × 60 columns
In [9]: df['Loan_Status_label'].value_counts(normalize=True)
Out[9]: Loan_Status_label
0 0.887346
1 0.112654
Name: proportion, dtype: float64
In [23]: df['Loan_Status_label'] = df['Loan_Status_label'].apply(lambda x: 1 if x == 'Fully Pai
X = df.drop('Loan_Status_label', axis=1)
y = df['Loan_Status_label']
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.2, stratify=y, random_state=42
)
print("Train shape:", X_train.shape)
print("Test shape:", X_test.shape)
Train shape: (32950, 59)
Test shape: (8238, 59)
In [25]: num_cols = X.select_dtypes(include=['int64', 'float64']).columns.tolist()
cat_cols = X.select_dtypes(include=['object']).columns.tolist()
numeric_pipeline = Pipeline([
('imputer', SimpleImputer(strategy='median')),
('scaler', StandardScaler())
])
categorical_pipeline = Pipeline([
('imputer', SimpleImputer(strategy='most_frequent')),
('encoder', OneHotEncoder(handle_unknown='ignore'))
])
preprocessor = ColumnTransformer([
('num', numeric_pipeline, num_cols),
('cat', categorical_pipeline, cat_cols)
])
In [27]: print("Train class distribution:\n", y_train.value_counts(normalize=True))
print("Test class distribution:\n", y_test.value_counts(normalize=True))
Train class distribution:
Loan_Status_label
0 1.0
Name: proportion, dtype: float64
Test class distribution:
Loan_Status_label
0 1.0
Name: proportion, dtype: float64
In [29]: from sklearn.utils import resample
X_fake = X.copy()
y_fake = y.copy()
if y_fake.nunique() == 1:
fake_positive = resample(X_fake, n_samples=int(len(X_fake) * 0.5), random_state
X_fake = pd.concat([X_fake, fake_positive])
y_fake = pd.concat([y_fake, pd.Series([1]*len(fake_positive))])
X_train, X_test, y_train, y_test = train_test_split(
X_fake, y_fake, test_size=0.2, stratify=y_fake, random_state=42
)
In [31]: final_model = Pipeline([
('pre', preprocessor),
('clf', RandomForestClassifier(random_state=42))
])
final_model.fit(X_train, y_train)
y_pred = final_model.predict(X_test)
y_proba = final_model.predict_proba(X_test)[:, 1]
# Evaluation
print("Accuracy:", accuracy_score(y_test, y_pred))
print("\nClassification Report:\n", classification_report(y_test, y_pred))
print("\nConfusion Matrix:\n", confusion_matrix(y_test, y_pred))
print("\nROC-AUC Score:", roc_auc_score(y_test, y_proba))
Accuracy: 0.2448814437161123
Classification Report:
precision recall f1-score support
0 0.39 0.22 0.28 8238
1 0.16 0.29 0.20 4119
accuracy 0.24 12357
macro avg 0.27 0.26 0.24 12357
weighted avg 0.31 0.24 0.26 12357
Confusion Matrix:
[[1850 6388]
[2943 1176]]
ROC-AUC Score: 0.2011164605829215
In [32]: # ROC Curve
fpr, tpr, thresholds = roc_curve(y_test, y_proba)
plt.figure(figsize=(6, 4))
plt.plot(fpr, tpr, label=f"ROC AUC = {roc_auc_score(y_test, y_proba):.2f}")
plt.plot([0, 1], [0, 1], 'k--')
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("ROC Curve")
plt.legend()
plt.grid(True)
plt.show()
In [ ]: X_train_transformed = preprocessor.fit_transform(X_train)
feature_names = preprocessor.get_feature_names_out()
rf = RandomForestClassifier(random_state=42)
rf.fit(X_train_transformed, y_train)
explainer = shap.TreeExplainer(rf)
shap_values = explainer.shap_values(X_train_transformed)
shap.summary_plot(shap_values[1], X_train_transformed, feature_names=feature_names
In [33]: plt.figure(figsize=(5,4))
sns.countplot(data=df, x='Loan_Status_label', palette='coolwarm')
plt.title("Loan Status Distribution")
plt.xlabel("Loan Status (0 = Charged Off, 1 = Fully Paid)")
plt.ylabel("Count")
plt.grid(axis='y')
plt.show()
In [37]: plt.figure(figsize=(12,8))
sns.heatmap(df.select_dtypes(include=['float64', 'int64']).corr(), annot=True,
plt.title("Correlation Heatmap")
plt.show()
In [41]: df['Loan_Status_label'].value_counts()
Out[41]: Loan_Status_label
0 41188
Name: count, dtype: int64
In [43]: import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
%matplotlib inline
# Minimal test DataFrame
df_test = pd.DataFrame({'Loan_Status_label': [0]*50 + [1]*30})
plt.figure(figsize=(5,4))
sns.countplot(data=df_test, x='Loan_Status_label', palette='coolwarm')
plt.title("Test Plot")
plt.xlabel("Loan Status")
plt.ylabel("Count")
plt.grid(axis='y')
plt.show()
In [51]: # Train Random Forest outside the pipeline
rf = RandomForestClassifier(random_state=42)
X_train_transformed = preprocessor.fit_transform(X_train) # Apply preprocessing
rf.fit(X_train_transformed, y_train) # Train model
Out[51]: ▾ RandomForestClassifier i ?
RandomForestClassifier(random_state=42)
In [61]: X_trans_df = pd.DataFrame(X_train_transformed, columns=preprocessor.get_feature_names_
X_trans_df['Loan_Status_label'] = y_train.values # Add target column
In [63]: top_features = feat_imp.head(5).index.tolist()
sample_df = X_trans_df[top_features + ['Loan_Status_label']].dropna().sample(200
sns.pairplot(sample_df, hue='Loan_Status_label', palette='Set1')
plt.suptitle("Pair Plot of Top Features", y=1.02)
plt.show()
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
