SakhiCircle

AI-Powered Credit Scoring Dashboard for Self-Help Groups (SHGs) in Rural India

Project Overview

SakhiCircle is a credit scoring platform designed for Self-Help Groups (SHGs) in rural India. It uses Machine Learning to predict creditworthiness based on group behavior and provides Explainable AI (SHAP) visualizations to show the "why" behind each score.

Key Features

• Credit Score (0-100) based on group financial metrics
• Random Forest ML Model trained on synthetic SHG data
• SHAP Waterfall Plots for explainable predictions
• MongoDB Database for data persistence
• Modern Dashboard UI with React + Tailwind CSS
• Role-Based Access Control (Admin/SHG Representative, Manager, User)
• Voice Input Support for loan purpose entry
• Score History & Trends with RadarChart and BarChart visualizations
• PDF Report Generation for loan applications

---

Novelty of the Proposed System

The key novelty of this system lies in its role-based financial intelligence framework, purpose-built for Self-Help Groups (SHGs). Unlike conventional SHG credit assessment mechanisms that rely heavily on external documentation, periodic audits, and manual verification, this platform embeds financial discipline, transparency, and accountability directly into daily SHG operations.

The system introduces three distinct user roles, each addressing a critical limitation in existing SHG–bank interaction models.

---

Member Role – Transparency and Financial Awareness

SHG members are provided with a personalized dashboard that displays:

• Meeting attendance records
• Monthly savings contributions
• Budget consistency indicators
• Overall group financial performance

By making these metrics visible to every member, the system promotes individual accountability and financial awareness. Members can clearly understand how their participation influences the group’s overall credit health, encouraging responsible and consistent financial behavior within the SHG.

---

Admin Role – Structured Data Capture and Internal Governance

The Admin role is assigned to a selected SHG member responsible for maintaining accurate group records. This role enables:

• Logging of meeting details and attendance
• Recording savings collections and cash flow
• Maintaining transaction and repayment data

Since this information forms the foundation for credit score computation, the Admin role ensures that financial data is structured, timely, and internally governed. This reduces dependence on paper-based records and minimizes inconsistencies that often result in loan delays or rejections.

---

Manager Role – Collective Loan Decision Interface

The Manager role serves as the formal interface between SHGs and lending institutions. Rather than individual loan applications, the SHG submits collective loan requests through this role.

An AI/ML-based evaluation engine analyzes:

• Attendance trends
• Savings discipline
• Repayment behavior
• Group stability indicators

Based on these factors, the system generates an explainable credit score and loan eligibility recommendation, supporting informed decision-making while preserving human oversight.

---

Why This Approach Is Necessary

Existing SHG lending systems suffer from:

• Fragmented and inconsistent records
• Limited transparency for members
• Subjective and manual credit evaluations

By integrating role-based data management, continuous financial monitoring, and explainable AI-driven credit assessment, this platform bridges the gap between informal SHG operations and formal financial institutions.

The system does not automate loan approvals. Instead, it functions as a decision-support framework that enhances trust, improves data quality, and strengthens financial readiness—making SHG lending fairer, more transparent, and scalable, while empowering SHGs to actively improve their own creditworthiness.

---

Loan Approval Workflow

The loan approval process follows a two-step verification system:

┌─────────────┐      ┌─────────────────┐      ┌─────────────┐      ┌──────────┐
│  SHG Member │ ──▶  │ SHG Representative │ ──▶  │   Manager   │ ──▶  │ Approved │
│   (User)    │      │     (Admin)       │      │             │      │   /Rejected
└─────────────┘      └─────────────────┘      └─────────────┘      └──────────┘
    Applies              Reviews &              Final
    for Loan             Forwards               Decision

Loan Status Flow:

1. Pending Admin Review - Member submitted, awaiting SHG Rep review
2. Pending Manager Review - Forwarded by SHG Rep, awaiting Manager decision
3. Approved - Loan approved by Manager
4. Rejected - Rejected by Manager
5. Rejected by Admin - Rejected by SHG Representative

---

Default Manager Credentials

• Username: manager
• Password: manager123

Important: Create users through registration page. Managers are pre-seeded in the database.

---

Tech Stack

Layer	Technology
Frontend	React 18 (Vite) + Tailwind CSS + Axios
Backend	Python FastAPI + Uvicorn
Database	MongoDB Atlas (Cloud)
AI/ML	Scikit-Learn (Random Forest) + SHAP
Charts	Recharts (RadarChart, BarChart)
Auth	JWT Token-based authentication

---

Project Structure

SakhiCircle/
├── backend/
│   ├── main.py              # FastAPI app with all endpoints
│   ├── model_trainer.py     # Script to generate data & train model
│   ├── requirements.txt     # Python dependencies
│   ├── shg_model.pkl        # Trained model (generated)
│   ├── shg_data.csv         # Sample data
│   └── .env                 # Environment variables (MongoDB URI)
│
├── frontend/
│   ├── src/
│   │   ├── App.jsx          # Main app with all dashboards
│   │   ├── main.jsx         # React entry point
│   │   ├── index.css        # Tailwind styles
│   │   └── components/
│   │       ├── ScoreCard.jsx   # Score display with gauge
│   │       ├── FormInput.jsx   # Slider input component
│   │       └── VoiceInput.jsx  # Voice-to-text input
│   ├── index.html
│   ├── package.json
│   ├── vite.config.js
│   ├── tailwind.config.js
│   └── postcss.config.js
│
└── README.md

---

Quick Start

Prerequisites

• Python 3.9+
• Node.js 18+
• MongoDB Atlas account (or local MongoDB)

Step 1: Setup Backend

# Navigate to backend
cd SakhiCircle/backend

# Create virtual environment
python -m venv venv

# Activate virtual environment
# Windows:
venv\Scripts\activate
# Mac/Linux:
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Train the ML model (IMPORTANT - run this first!)
python model_trainer.py

# Start the API server
python main.py
# OR
uvicorn main:app --reload --port 8000

The backend will be available at: http://localhost:8000

• API Docs: http://localhost:8000/docs
• Health Check: http://localhost:8000/health

On first startup, a default admin user is created automatically.

Step 2: Setup Frontend

# Open new terminal, navigate to frontend
cd SakhiCircle/frontend

# Install dependencies
npm install

# Start development server
npm run dev

The frontend will be available at: http://localhost:3000

---

API Endpoints

Authentication Endpoints

Endpoint	Method	Role	Description
/login	POST	Public	Login and get auth token
/register	POST	Public	Register new user (Admin/User)
/logout	POST	Any	Invalidate auth token

Loan Endpoints

Endpoint	Method	Role	Description
/loan/apply	POST	User	Submit loan application (status: Pending Admin Review)
/loan/my_requests	GET	User	View own applications
/loan/pending_admin_review	GET	Admin	Get requests pending admin review
/loan/forward_to_manager	POST	Admin	Forward request to manager
/loan/reject_by_admin	POST	Admin	Reject request without forwarding
/loan/all	GET	Manager	Get all requests pending manager review
/loan/update_status	POST	Manager	Approve/Reject request
/loan/history	GET	Manager/Admin	View all loan history

SHG Group Endpoints

Endpoint	Method	Role	Description
/shg/group_data	GET	Admin	Get group financial data
/shg/update_group_data	POST	Admin	Update group metrics
/shg/my_group_data	GET	User	Get group data for members
/score/log	POST	Admin	Log score to history
/score/history/{shg_name}	GET	Manager	Get score history for trend analysis

Prediction Endpoints

Endpoint	Method	Description
/predict	POST	Calculate credit score with SHAP explanation
/health	GET	Check API and model status

POST /predict

Predict credit score for an SHG.

Request:

{
  "savings": 2500,
  "attendance": 85,
  "repayment": 75
}

Response:

{
  "score": 78,
  "risk": "Low Risk ✅",
  "risk_color": "#22c55e",
  "explanation_image": "data:image/png;base64,...",
  "features": {
    "savings_per_member": 2500,
    "attendance_rate": 85,
    "loan_repayment_rate": 75,
    "low_risk_probability": 0.92
  },
  "timestamp": "2026-01-07T10:30:00.000Z"
}

POST /login

Authenticate user.

Request:

{
  "username": "admin",
  "password": "admin123"
}

Response:

{
  "success": true,
  "token": "your-auth-token",
  "role": "admin",
  "username": "admin",
  "user": {
    "id": "...",
    "full_name": "System Administrator",
    "role": "admin",
    "branch_id": "HQ"
  }
}

---

ML Model Details

Features

Feature	Range	Weight
Savings per Member	₹100 - ₹5,000	30%
Attendance Rate	0% - 100%	30%
Loan Repayment Rate	0% - 100%	40%

Credit Score Formula

Score = 0.30 × Normalized_Savings + 0.30 × Attendance + 0.40 × Repayment

Risk Classification

• Low Risk: Score ≥ 60
• High Risk: Score < 60

---

Explainable AI

SakhiCircle uses SHAP (SHapley Additive exPlanations) to explain predictions:

• Each prediction includes a Waterfall Plot
• Shows how each feature pushed the prediction up or down
• Makes AI decisions transparent and trustworthy

---

UI Features

Login Page

• Single entry point for all users
• No public signup (admin creates users)
• Auto-redirect based on role

Admin Panel

• Create/delete users
• Assign roles (admin, manager, user)
• Set branch IDs and KYC status

Manager Inbox

• View pending loan requests
• Download PDF reports with SHAP charts
• Approve or reject applications

SHG Dashboard

• Calculate credit score
• View SHAP explanation
• Apply for loans (one pending at a time)
• Track application status

---

MongoDB Setup

MongoDB is required for this application:

1. Install MongoDB locally or use MongoDB Atlas
2. The default connection uses MongoDB Atlas (configured in code)
3. To use a custom MongoDB:

   set MONGO_URI=mongodb://localhost:27017

4. Restart the backend

Collections Used

• users - User accounts with roles
• loan_requests - Loan applications
• auth_tokens - Authentication tokens
• shg_logs - Prediction audit logs

---

Video Link

https://vimeo.com/1152687391?fl=ip&fe=ec

---

License

MIT License - Feel free to use and modify!

---

Acknowledgments

• Self-Help Group movement in India
• SHAP library for Explainable AI
• FastAPI & React communities