import pandas as pd  

# Load dataset  
df = pd.read_csv('loan_recovery_dataset.csv')  

# Display basic info  
print(df.info())  
print(df.head())  

# Fill missing values for numerical columns with median  
df.fillna({'Monthly_Income': df['Monthly_Income'].median(),  
           'Collateral_Value': df['Collateral_Value'].median(),  
           'Outstanding_Loan_Amount': df['Outstanding_Loan_Amount'].median(),  
           'Num_Missed_Payments': 0,  
           'Days_Past_Due': 0}, inplace=True)  

# Fill categorical missing values with mode  
df.fillna({'Employment_Type': df['Employment_Type'].mode()[0],  
           'Collection_Method': df['Collection_Method'].mode()[0]}, inplace=True)  

# Convert categorical variables to numerical  
df['Employment_Type'] = df['Employment_Type'].map({'Salaried': 1, 'Self-Employed': 0})  
df['Gender'] = df['Gender'].map({'Male': 1, 'Female': 0})  

import matplotlib.pyplot as plt  
import seaborn as sns  

# Loan Recovery Distribution  
sns.countplot(x='Recovery_Status', data=df, palette='coolwarm')  
plt.title('Loan Recovery Status')  
plt.show()  

plt.figure(figsize=(12, 6))  
sns.heatmap(df.corr(), annot=True, cmap='coolwarm')  
plt.title('Feature Correlation')  
plt.show()  

# Select only numerical features for correlation analysis
numerical_features = df.select_dtypes(include=np.number).columns
correlation_matrix = df[numerical_features].corr()

plt.figure(figsize=(12, 6))
sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm')
plt.title('Feature Correlation')
plt.show()

# Debt-to-Income Ratio  
df['Debt_to_Income'] = df['Outstanding_Loan_Amount'] / df['Monthly_Income']  

# Loan Risk Score (Custom Formula)  
df['Loan_Risk_Score'] = df['Interest_Rate'] * df['Num_Missed_Payments'] / df['Loan_Amount']  

from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, classification_report
from sklearn.preprocessing import LabelEncoder # Import LabelEncoder

# Define features (X) and target (y)
X = df.drop(['Borrower_ID','Legal_Action_Taken' , 'Loan_ID', 'Recovery_Status'], axis=1)
y = df['Recovery_Status']

# Convert 'Payment_History' to numerical using Label Encoding before one-hot encoding
le = LabelEncoder() # Create a LabelEncoder object
X['Payment_History'] = le.fit_transform(X['Payment_History']) # Fit and transform the column

# Convert categorical features to numerical using one-hot encoding
categorical_features = ['Loan_Type', 'Collection_Method', 'Payment_History']
X = pd.get_dummies(X, columns=categorical_features, drop_first=True)

# Split into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train RandomForest Model
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train, y_train)

# Make predictions
y_pred = model.predict(X_test)

# Evaluate Model
print("Accuracy:", accuracy_score(y_test, y_pred))
print(classification_report(y_test, y_pred))

import joblib  

# Save the trained model  
joblib.dump(model, 'loan_recovery_model.pkl')  
print("Model saved successfully!")