A Review on Hybrid Machine Learning and Deep Learning Approaches for Fraud Detection in Online Marketplace Transactions

Abstract

E-commerce platforms have expanded rapidly over past decades, resulting in significantly increased online transactions and increased fraudulent activities. Traditional techniques in fraud detection, which include rule-based systems and conventional machine learning models, struggle to accommodate high-dimensional data, swift evolving fraud patterns, or severe class imbalance. The study offers an extensive review of the existing approaches in fraud detection, focusing on the hybrid approaches of machine learning and deep learning. Particular emphasis is given to the use of advanced pre-processing methods, such as data normalization, categorical encoding, and Synthetic Minority Over-sampling Technique (SMOTE), to deal with data imbalance problems. The study also observes and evaluates different algorithms like Logistic Regression, Decision Trees, Random Forest, Gradient Boosting, and CatBoost, plus deep learning techniques-such as Autoencoders-in detecting anomalies. Another point to note is the importance of hybrid approaches utilizing both supervised and unsupervised learning to improve both performance accuracy and model robustness. The study also indicates that CatBoost is good at handling categorical features, while also restricting overfitting; Autoencoders, due to their reconstruction errors, tend to effectively capture stealth anomalies. The research also elaborates on key evaluation metrics, sufficient accuracy, precision, recall, and F1-score, to assess the performance of models for imbalanced datasets. It argues that hybrid methods do much better than traditional technique performances by enhancing the sensitivity of fraud detection while reducing false negatives. Finally, paper ends with future research trends Danstellerity defines the future trends in this discussion by integrating explainable AI and superior deep learning architectures, which would create significant fraud detection systems on a larger scale.

How to cite this article:
Mishra M, Rai A K. A Review on Hybrid Machine Learning and Deep Learning Approaches for Fraud Detection in Online Marketplace Transactions. J Adv Res Comp Tech Soft Appl 2026; 10(2): 1-9.