Fraud Detection on e-Commerce using Machine Learning Techniques

Abstract

Fraud detection holds significant importance in e-commerce transactions as it serves unauthorized activities like identity theft, account takeovers, and fraudulent transactions. Recently, machine learning algorithms have gained widespread adoption for fraud detection in e-commerce transactions. These algorithms function by discerning patterns within the data indicative of fraudulent behavior. Pattern detection entails identifying discriminative features in the data, such as irregular transaction amounts, locations, or behaviors deviating from a user's norm, by inputting into the machine learning model. In this research, four fundamental machine learning algorithms (decision tree, random forest, voting classifier) are employed for fraud detection in e-commerce transactions, utilizing a freshly curated dataset containing diverse features related to online shopping activities on Boyner Group's e-commerce platform and mobile app. This study contributes to the existing literature by experimenting with various machine learning classifiers and incorporating distinct features, diverging from prevailing methodologies in the field.