FinLend

FinLend is a loan recommendation prototype that combines a Laravel web interface with a FastAPI machine-learning engine. It predicts borrower risk and uses fuzzy logic to recommend loan status, credit limit, and interest rate.

Overview

The project is split into two main parts:

• finlend-web: Laravel application for the user-facing credit form and results page.
• VSC: FastAPI AI engine that loads trained model artifacts and exposes a credit calculation endpoint.

The Laravel app collects applicant data, builds the expected feature vector, and sends it to the Python API. The Python API combines a neural-network risk prediction with fuzzy decision rules for final credit recommendations.

Key Features

• Credit application form built in Laravel.
• FastAPI endpoint for model-backed credit calculation.
• MLP-based default-risk prediction using saved model, scaler, encoders, and feature columns.
• Fuzzy logic engine for loan limit, interest, and approval status.
• Validation for feature count and income limits.
• Result page showing risk and recommended credit decision.
• Separation between web UI and AI inference service.

Tech Stack

Web App

• PHP 8.2
• Laravel 12
• Vite
• Tailwind CSS 4

AI Engine

• Python
• FastAPI
• Uvicorn
• pandas
• numpy
• scikit-learn
• scikit-fuzzy
• joblib

Project Structure

finlend-web/                  Laravel web application
finlend-web/app/Http/         Controllers and web logic
finlend-web/routes/web.php    Web routes for form and calculation
VSC/main.py                   FastAPI application
VSC/fuzzy_engine.py           Fuzzy credit decision rules
VSC/*.pkl                     Saved model, scaler, encoders, and feature metadata

Getting Started

Run the AI engine first:

cd VSC
python -m venv .venv
.venv\Scripts\activate
pip install fastapi uvicorn pandas numpy scikit-learn scikit-fuzzy joblib
uvicorn main:app --reload --host 127.0.0.1 --port 8000

Run the Laravel web app in a second terminal:

cd finlend-web
composer install
npm install
copy .env.example .env
php artisan key:generate
php artisan serve
npm run dev

By default, the Laravel controller posts to:

http://127.0.0.1:8000/hitung-kredit

API Summary

POST /hitung-kredit
Content-Type: application/json

{
  "features": [/* ordered numeric/categorical feature vector */]
}

The response contains model risk output and fuzzy decision output.

Limitations

• The model depends on the exact feature order and preprocessing artifacts used during training.
• Income and loan limit values follow the dataset/model scale. Keep currency conversion consistent if the UI displays IDR.
• The recommendation should be treated as decision support, not an automatic loan approval system.
• Real lending use would require explainability, bias testing, compliance review, audit logging, and manual override workflows.

Suggested Tests

• Contract tests between Laravel feature vector creation and FastAPI expected columns.
• API tests for invalid feature length, invalid income, and missing model artifacts.
• Regression tests for fuzzy rule changes.
• Evaluation against holdout data before presenting the model as reliable.

Status

Academic/portfolio fintech prototype demonstrating web-to-AI integration for credit risk decision support.