💳 Finora — AI-Powered Microfinance Intelligence Platform

Finora is an enterprise-grade Machine Learning platform designed to predict microfinance loan repayment behavior using telecom customer data. The system helps financial institutions identify potential defaulters, reduce lending risk, and improve credit decision-making through AI-driven analytics.

---

📌 Problem Statement

Microfinance Institutions (MFIs) provide financial services such as small loans, agricultural loans, and business loans to low-income and underserved populations. These services are especially valuable in remote areas where traditional banking access is limited.

A telecom company collaborating with an MFI launched a micro-credit service that allows customers to borrow mobile balance amounts that must be repaid within 5 days.

Loan repayment structure:

• Loan Amount: 5 → Repayment: 6
• Loan Amount: 10 → Repayment: 12

Customers who fail to repay within the repayment period are considered defaulters.

The challenge is to build a predictive Machine Learning model capable of identifying whether a customer is likely to repay the loan successfully.

---

🎯 Objective

The primary objective of this project is to:

• Predict loan repayment probability for telecom microfinance customers
• Identify potential defaulters using Machine Learning
• Improve customer selection for micro-credit lending
• Reduce financial risk for telecom operators and MFIs
• Provide intelligent AI-driven credit risk analytics

Target Labels:

• 1 → Non Defaulter (Loan Repaid)
• 0 → Defaulter (Loan Not Repaid)

---

🚀 Features

• 📊 Advanced Loan Repayment Prediction
• 🤖 Machine Learning Risk Classification
• 📈 Interactive Fintech Dashboard
• 💳 Customer Risk Analytics
• 📁 CSV Upload & Batch Prediction
• 📉 Defaulter Detection System
• 📥 Downloadable Prediction Reports
• 🎨 Modern Enterprise Streamlit UI

---

🧠 Machine Learning Workflow

The project follows a complete end-to-end ML pipeline:

1. Data Cleaning
2. Exploratory Data Analysis (EDA)
3. Feature Engineering
4. Data Preprocessing
5. Model Training
6. Hyperparameter Tuning
7. Model Evaluation
8. Deployment with Streamlit

---

📊 Evaluation Metrics

Models were evaluated using:

• Log Loss
• Precision
• Recall
• F1 Score
• ROC-AUC Score

---

🛠️ Tech Stack

Frontend

• Streamlit
• Custom CSS

Machine Learning

• Scikit-Learn
• Random Forest
• XGBoost
• LightGBM
• CatBoost

Data Processing

• Pandas
• NumPy

Visualization

• Matplotlib
• Seaborn

---

📂 Project Structure

Finora/
│
├── app.py
├── README.md
├── requirements.txt
│
├── data/
│
├── notebooks/
│   └── loan_prediction.ipynb
│
├── outputs/
│   ├── best_model.pkl
│   ├── submission.csv
│   └── plots/
│
└── src/
    ├── preprocessing.py
    ├── feature_engineering.py
    └── modeling.py