Fraud Detection Flask ML Project

This repository contains a complete end-to-end Machine Learning project for detecting fraudulent insurance claims using a Flask-based web application. The project includes data ingestion, validation, preprocessing, clustering, model training (with multiple algorithms including XGBoost), model selection, and prediction via a web interface.

Project Overview

This is a fraud detection system built for insurance claims data. The pipeline handles:

• Raw data validation
• Data preprocessing (handling missing values, feature engineering)
• Clustering (K-Means for grouping similar data)
• Training multiple ML models
• Selecting the best model using cross-validation
• Saving the trained model
• Batch and single prediction via Flask API
• Monitoring dashboard integration

The application runs locally and provides a simple web form to upload CSV files for prediction.

What Was Done (Project Setup & Deployment Steps)

Here is a step-by-step summary of what was performed to set up and run this project (as executed on January 3, 2026):

1. Navigated to Project Directory

   • cd Desktop
     cd fraud

2. Created a Dedicated Conda Environment

   • Accepted Anaconda Terms of Service for required channels.
   • Created a new environment with Python 3.8:

     conda create -n fraud-env python=3.8 -y
     conda activate fraud-env

3. Installed Dependencies

   • Installed all required packages from requirements.txt (includes Flask, scikit-learn, XGBoost, pandas, numpy, Flask-MonitoringDashboard, etc.):

     pip install -r requirements.txt

4. Initialized Git Repository

   git init
   git add .
   git commit -m "Initial commit - Fraud detection Flask ML project"
   git branch -M main

5. Pushed to GitHub

   • Added remote origin and pushed the code:

     git remote add origin https://github.com/sujith52/fraud.git
     git push -u origin main

6. Ran the Flask Application

   python main.py

   • Started the server on http://127.0.0.1:5001
   • Scheduler and monitoring dashboard initialized
   • Successfully served the main page and handled prediction requests via /predict endpoint

How to Run the Project

Prerequisites

• Miniconda/Anaconda installed
• Git installed

Setup Instructions

1. Clone the repository:

   git clone https://github.com/sujith52/fraud.git
   cd fraud

2. Create and activate conda environment:

   conda create -n fraud-env python=3.8 -y
   conda activate fraud-env

3. Install dependencies:

   pip install -r requirements.txt

4. Run the application:

   python main.py

5. Open browser and go to:

   http://127.0.0.1:5001
   

6. Use the web form to upload a CSV file (in the expected format) for fraud prediction.

Monitoring Dashboard

• Access performance monitoring at: http://127.0.0.1:5001/dashboard
• Uses Flask-MonitoringDashboard for tracking API usage and performance.

Project Structure

Key directories and files:

• data/ - Contains training dataset (insuranceFraud.csv)
• models/ - Stores trained model files
• main.py - Entry point for Flask app
• predictFromModel.py - Prediction logic
• trainingModel.py - Model training pipeline
• templates/index.html - Web interface
• requirements.txt - All Python dependencies (pinned to compatible versions)
• schema_training.json / schema_prediction.json - Data schema validation rules

Dataset

The model was trained on insurance claims data with features leading to a target column indicating fraud (fraud_reported or similar).

Technologies Used

• Python 3.8
• Flask (Web Framework)
• scikit-learn, XGBoost (ML Models)
• Pandas, NumPy (Data Processing)
• K-Means Clustering (Data Grouping)
• Flask-MonitoringDashboard (App Monitoring)
• Git & GitHub (Version Control)

Notes

• This project uses older versions of libraries (e.g., Flask 1.1.1, scikit-learn 0.22.1) for compatibility.
• For production, use Gunicorn + Nginx instead of Flask's development server.
• The Procfile suggests readiness for deployment on platforms like Heroku.

Author

@sujith52

Feel free to fork, improve, or raise issues!