tozenz/loan-approval-model

Loan Approval Prediction Model

Model Details

• Model Architecture: Logistic Regression
• Task: Binary Classification (Loan Approval/Rejection)
• Framework: scikit-learn
• Training Data: Loan dataset with customer financial information

Intended Use

This model predicts whether a loan application should be approved or rejected based on customer financial and personal information.

Input Features

The model expects 13 numerical features:

• Gender, Married, Dependents, Education, Self_Employed (encoded)
• ApplicantIncome, CoapplicantIncome, LoanAmount
• Loan_Amount_Term, Credit_History, Property_Area (encoded)
• Age (standardized)

Output

• 0: Loan Rejected
• 1: Loan Approved

How to Use

In Python

import joblib
import numpy as np

# Load model and scaler
model = joblib.load('loan_model.pkl')
scaler = joblib.load('scaler.pkl')

# Prepare your features (13 features in correct order)
# Features should be in the same order as training data
features = np.array([[1, 0, 0, 1, 0, 5000, 0, 100, 360, 1, 2, 1, 30]])

# Scale features
features_scaled = scaler.transform(features)

# Make prediction
prediction = model.predict(features_scaled)[0]
probability = model.predict_proba(features_scaled)[0]

print(f"Prediction: {'Approved' if prediction == 1 else 'Rejected'}")
print(f"Approval Probability: {probability[1]:.2%}")

Via API (using website)

The model is integrated into a Flask/React web application that:

1. Takes user input through an intuitive form
2. Preprocesses and scales the features
3. Returns prediction with confidence score
4. Displays loan decision and key factors

Model Performance

• Accuracy: ~80-82%
• Precision: ~82-85%
• Recall: ~75-78%
• F1-Score: ~78-81%

Technical Details

Preprocessing

• Categorical encoding (Gender, Married, Education, Self_Employed, Property_Area)
• Standard scaling (StandardScaler) for numerical features
• Missing value handling and data cleaning applied

Training

• Algorithm: Logistic Regression with L2 regularization (default)
• Max iterations: 1000
• Random seed: 42 (for reproducibility)

Model Card Author

Data Science Student - Year 3, Term 2

Intended Users

• Loan approval decision automation
• Credit risk assessment
• Educational purposes

Limitations

• Model trained on specific historical loan data
• Performance may vary on different data distributions
• Should not be sole basis for loan decisions
• Consider domain expertise and additional factors

License

MIT License - Free to use and modify

---

Repository Structure:

• loan_model.pkl - Trained logistic regression model
• scaler.pkl - Feature scaler for preprocessing
• README.md - This model card