⇒ Loan Approval Prediction System

A production-ready machine learning system that predicts loan approval status using ensemble methods (Random Forest, XGBoost, LightGBM) with MLflow experiment tracking, FastAPI backend, and an interactive web interface.

⇒ Quick Start

- Automated Setup (Recommended)

Run the automated setup script that handles environment creation, dependencies, MLflow, and API:

bash run.sh

This interactive script will:

1. ✅ Create/activate conda environment
2. ✅ Install dependencies from requirements.txt
3. ✅ Start MLflow UI (http://localhost:5000)
4. ✅ Optionally run model training
5. ✅ Start FastAPI server (http://localhost:8000)

⇒ Installation

- Method 1️⃣: Local Setup

Step 1: Create Conda Environment

conda create -n deploy python=3.9.23
conda activate deploy

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Start MLflow UI

Open a terminal and run:

mlflow ui

🔗 Access MLflow at: http://127.0.0.1:5000

Step 5: Train Model (Optional)

python src/test_1.py

With custom parameters:

python src/test_1.py -n 300 -xlr 0.05 -llr 0.07

Step 6: Start FastAPI Server

In a new terminal:

conda activate deploy
python src/api_1.py

🔗 API : http://127.0.0.1:8000 🔗 API Doc: http://127.0.0.1:8000/docs

Step 7: Open Frontend

Simply open frontend/index.html in your browser.

---

- Method 2️⃣: Docker

Step 1: Build Docker Image

docker build -t loan_app .

Step 2: Run Container

docker run -p 8000:8000 -p 5000:5000 loan_app

Exposed Ports:

• 8000: FastAPI server
• 5000: MLflow UI (optional)

Step 3: Access Services

• API Docs: http://127.0.0.1:8000/docs
• Frontend: Open frontend/index.html in your browser

Stop & Remove Container

# List running containers
docker ps

# Stop container
docker stop <container_id>

# Remove container
docker rm <container_id>

🗂️ Project Structure

Loan_Approval_Prediction/
├── data/                              # Dataset directory
│   └── loan_approval_dataset.csv
├── frontend/                          # Web interface
│   ├── index.html                     # Main HTML page
│   └── main.js                        # Frontend JavaScript logic
├── notebook/                          # Exploratory analysis
│   └── test1.ipynb                    # EDA & model experimentation
├── output/                            # Generated artifacts
│   ├── best_loan_model.pkl            # Trained model
│   └── model_results.csv              # Model comparison results
├── src/                               # Source code
│   ├── api_1.py                       # FastAPI application
│   └── test_1.py                      # Model training script
├── Dockerfile                         # Docker configuration
├── helperFunction.py                  # Utility functions
├── README.md                          # Project documentation
├── requirements.txt                   # Python dependencies
└── run.sh                             # Automated setup script

⇒ Model Training

Custom Training Parameters

python src/test_1.py -n <n_estimators> -xlr <xgb_learning_rate> -llr <lgbm_learning_rate>

Example:

python src/test_1.py -n 300 -xlr 0.05 -llr 0.07

Tracked Metrics

• Accuracy (Train & Test)
• Balanced Accuracy
• F1 Score (Macro & Micro)
• ROC-AUC Score
• Accuracy Gap (overfitting indicator)
• Training Time

.. This is for DEPI Deployment task ..

THANK YOU FOR VISITING!