Application Of Machine Learning To Detect Credit Card Fraud Transactions In E-Commerce: A Case Study Of Abc Solutions Company

Abstract

Fraud detection in e-commerce transactions is a critical challenge that demands effective and reliable solutions to protect businesses and consumers from financial losses and security breaches. In this study, we propose a novel approach based on the Prudential Multiple Consensus (PMC) model, incorporating feature engineering, feature selection, and resampling techniques, to address fraud detection in e-commerce settings. The PMC model intelligently combines multiple classification algorithms using a prudential criterion, optimizing their performance in identifying fraudulent activities. Additionally, we compare the PMC model with various ensemble approaches, including complete agreement, majority voting, weighted voting, classifier selection, and pairwise accuracy, while utilizing a comprehensive set of algorithms. Our experiments demonstrate that the PMC model, in conjunction with feature engineering, feature selection, and resampling, outperforms these alternative strategies, exhibiting superior fraud detection accuracy. By employing time series cross-validation and considering the chronological order of transactions, our approach proves to be particularly effective in capturing temporal fraud patterns. The results highlight the PMC model's efficacy and its potential to serve as a robust and reliable solution for fraud detection in e-commerce transactions, contributing to enhanced security and trust in online commerce