https://github.com/samkazan/fraud-detection-ml

Machine learning models for enhanced fraud detection in e-commerce transactions, exploring feature engineering, distance prediction, and clustering analysis.
https://github.com/samkazan/fraud-detection-ml

clustering data-science data-visualization dataanalytics dbscan eda hierarchical-clustering kmeans-clustering knn-imputer matplotlib mlxtend python scikit-learn seaborn xgboost

Last synced: about 16 hours ago
JSON representation

Machine learning models for enhanced fraud detection in e-commerce transactions, exploring feature engineering, distance prediction, and clustering analysis.

• Host: GitHub
• URL: https://github.com/samkazan/fraud-detection-ml
• Owner: SamKazan
• Created: 2024-06-27T01:35:17.000Z (4 months ago)
• Default Branch: main
• Last Pushed: 2024-06-27T04:19:43.000Z (4 months ago)
• Last Synced: 2024-06-27T05:55:09.143Z (4 months ago)
• Topics: clustering, data-science, data-visualization, dataanalytics, dbscan, eda, hierarchical-clustering, kmeans-clustering, knn-imputer, matplotlib, mlxtend, python, scikit-learn, seaborn, xgboost
• Language: Jupyter Notebook
• Homepage:
• Size: 11.4 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

README

        
# Machine Learning for Fraud Detection in E-commerce Transactions

## Overview

This project investigates the application of machine learning techniques to enhance fraud detection in e-commerce transactions. By leveraging a comprehensive dataset from Vesta, we explore feature engineering, distance prediction, and clustering analysis to identify fraudulent activities.

## Problem Statement

The increasing sophistication of financial fraud poses significant challenges to businesses and consumers. Traditional rule-based fraud detection systems often struggle to keep pace with evolving fraudulent tactics. This project aims to develop more robust and accurate fraud detection models using machine learning.

## Methodology

This project addresses three key research questions:

**RQ1: Feature Engineering and Selection**

* **Objective:** Improve fraud detection accuracy by identifying and engineering the most predictive features.

* **Techniques:** Recursive Feature Elimination (RFE), Feature Importance from Gradient Boosting, Principal Component Analysis (PCA).

**RQ2: Predicting Transaction Distances**

* **Objective:** Develop models to predict transaction distances and identify geographic anomalies indicative of fraud.

* **Techniques:** Linear Regression, XGBoost.

**RQ3: Clustering for Coordinated Fraud Detection**

* **Objective:** Utilize clustering techniques to uncover groups of transactions potentially associated with coordinated fraud.

* **Techniques:** K-Means Clustering, HDBSCAN, Hierarchical Clustering.

## Results

* **Feature Engineering:** PCA significantly enhanced model accuracy, highlighting its effectiveness in capturing relevant data structures.

* **Distance Prediction:** XGBoost models demonstrated promising results in predicting transaction distances, aiding in the identification of high-risk transactions.

* **Clustering Analysis:** K-Means Clustering provided the most interpretable and well-separated clusters, potentially revealing patterns of coordinated fraud.

## Data

* **Source:** "IEEE-CIS Fraud Detection" dataset from Kaggle, provided by Vesta.

* **Size:** Over 140,000 transactions with 434 features (transaction details, card information, addresses, Vesta-engineered features).

## Contact

Cem Kazan - [email protected]# fraud-detection-ml