https://github.com/yuberlinets/transactions-fraud

Fraudulent Transaction detection
https://github.com/yuberlinets/transactions-fraud

analysis docker-compose fastapi fraud-detection java kafka kafka-streams ml mongodb nginx python react real-time rest-api sockjs spring spring-boot typescript websocket

Last synced: 2 months ago
JSON representation

Fraudulent Transaction detection

• Host: GitHub
• URL: https://github.com/yuberlinets/transactions-fraud
• Owner: YuBerlinets
• License: apache-2.0
• Created: 2025-02-03T22:15:11.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-02-08T20:17:02.000Z (over 1 year ago)
• Last Synced: 2025-04-09T05:44:18.386Z (about 1 year ago)
• Topics: analysis, docker-compose, fastapi, fraud-detection, java, kafka, kafka-streams, ml, mongodb, nginx, python, react, real-time, rest-api, sockjs, spring, spring-boot, typescript, websocket
• Language: Java
• Homepage:
• Size: 386 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Transaction Fraud Detections

### Real-time analysis of banking transactions to detect fraudulent ones

## Overview

This project is a real-time fraud detection system for banking transactions. It utilizes Apache Kafka for event streaming, a machine learning model trained with DecisionTreeClassifier from sklearn, and a Spring Boot backend to process transactions. Fraudulent transactions are stored in MongoDB, while legitimate ones (in a real-world scenario) would be forwarded to an internal banking system.

## Architecture

![Architecture Scheme](https://github.com/user-attachments/assets/8a438fcb-6e7e-4ef4-b6e4-bdb537993a29)

### Transaction Producer (Python)

- Generates 100 transactions per minute for testing.

- Sends transactions to the Kafka topic transactions.

### Backend (Java Spring Boot)

- Consumes transactions from Kafka.

- Calls an ML-based fraud detection model (DecisionTreeClassifier).

- If fraudulent, stores the transaction in MongoDB and sends to "fraud-alert" Kafka topic

- If legitimate, in a real-world case, forwards the transaction to a banking system.

- Sends transaction updates via WebSocket to the frontend.

  

### Frontend (React + TypeScript)

- Real-time dashboard displaying: Transactions per 3 minutes, Blocked Transactions, Approved Transactions, Fraud Alerts

- HTTP server configured using Nginx

### Transactions Streaming (Kafka)

- Has two topics with events (fraud-alerts, transactions)

### ML Model (Python + FastAPI + SkLearn)

- Trained on datasets with transaction samples based on the Decision Tree Classifier

- Has endpoints to send and analyze transactions and evaluate a model

## Tech Stack

- **Backend**: Java, Spring Boot, Kafka Streams, MongoDB, FastAPI, WEBSocket

- **Machine Learning**: Python, Scikit-learn

- **Frontend**: React, TypeScript, Vite, SockJS, Nginx

- **Event Streaming**: Apache Kafka

## Client Layout

**Dashboard**

![Dashboard Light](https://github.com/user-attachments/assets/7a507fe6-de1f-4b07-8c00-9bda76e5a544)  

![Dashboard Dark](https://github.com/user-attachments/assets/a80b4f1d-9fee-46ba-8326-0d07debc796b)  

**Stored Fraudulent Transactions**

![DB Light](https://github.com/user-attachments/assets/d530af8b-1fd5-4c39-b0e3-2362523442de)

![DB Dark](https://github.com/user-attachments/assets/4b7d03e8-7cb0-4ba4-b428-c0ea107666d4)

## Config (.env files)

### **Client**  

VITE_BASE_URL=  

  

### **Java Backend**

MONGODB_URI=  

KAFKA_BOOTSTRAP_SERVERS=  

ML_MODEL_URL=  

KAFKA_STREAMS_APPLICATION_ID=  

BOOTSTRAP_SERVERS_CONFIG=  

  

### **Python Generator**  

BASE_URL=  

## How to run?

> [!WARNING]

Don't forget to change the .env files and properties in docker-compose.yml!  

To run using Docker compose run this command in project folder.  

```

docker-compose up --build

```