fraud-detection-ML-pipeline

A comprehensive end-to-end machine learning pipeline for credit card fraud detection with high accuracy (~99.6%). This project includes the complete workflow from data exploration to model deployment with an integrated CI/CD pipeline.

Project Overview

This project implements a machine learning system to detect fraudulent credit card transactions. It uses a custom-built fraud detection model trained on transaction data that includes temporal patterns, merchant information, and customer demographics.

Key features of this pipeline:

• Data preprocessing with custom transformations for categorical and temporal features
• Exploratory data analysis revealing insights into fraud patterns
• Model training and evaluation with high performance metrics
• Complete CI pipeline for automated testing and deployment
• Comprehensive logging system for tracking model performance and issues
• Visualization tools for understanding model decisions and fraud patterns

Key Insights from Data Analysis

Our exploratory data analysis revealed several important patterns:

1. Age-Based Vulnerability: People over 50 years old tend to be more vulnerable to fraud compared to younger age groups.

2. Temporal Patterns: Fraud rates vary significantly by hour of day and day of week, with late night - early morning showing higher risk.

3. Geographic Hotspots: Large cities like Washington, New York, and Los Angeles have the highest number of fraudulent transactions.

4. Transaction Categories: Shopping and groceries are the transaction types showing higher fraud rates than others.

Installation

1. Clone the Repository:

   git clone https://github.com/ol1g3/fraud-detection-ML-pipeline.git
   cd fraud-detection-ML-pipeline

2. Set Up the Virtual Environment: Choose one of the methods below.

Virtual Environment Setup Options

Using venv (default)

1. Create the Virtual Environment (Mac):

   python3 -m venv .venv

2. Activate the Virtual Environment:

   source .venv/bin/activate
   pip install -r requirements.txt

3. Deactivate the Virtual Environment (when done):

   deactivate

Using uv (alternative)

1. Create the Virtual Environment:

   uv venv --python 3.11

2. Activate the Virtual Environment And Install Requirements:

   source .venv/bin/activate
   uv pip install -r requirements.txt

3. Deactivate the Virtual Environment (when done):

   deactivate

Automated CI Pipeline

This project includes a complete CI/CT pipeline that:

• Runs automated tests on every commit, including build, unit tests
• Validates model performance on test data (regression test)
• Generates performance reports and logs

Future improvements

• Implement neural network-based approaches for potentially higher accuracy
• Add more sophisticated feature engineering based on domain knowledge

Data Source

The dataset used for this project can be found on Kaggle: Fraud Detection Dataset.