• V Divya Jyothi
• Dr Y.A.Siva Prasad

The detection and prevention of fraud ulent transactions on e-commerce platforms remain critical aspects of ensuring transaction security. How ever, the inherently covert nature of e-commerce activities makes it challenging to identify malicious actors based solely on historical transaction data. Existing research efforts to often overlook the dy namic behavioral patterns of users from multiple dimensions, resulting with suboptimal fraud detec tion performance. This paper integrates Risk-Based Approach (RBA) and Deep Neural Network (DNN) techniques by incorporating internal control risk in dicators alongside traditional Anti-Money Laundering (AML) algorithms. Using Proof of Concept (POC) data for model evaluation, the Autoencoder (AE) was identified as the most effective unsupervised learning model for this task. The developed predictive model focuses on accurately identifying fraudulent activity in previously unseen data, emphasizing improved generalization capabilities. To mitigate overfitting caused by hyperparameter configurations tightly aligned with the training dataset, dropout regularization was implemented during model training. This approach en hances the robustness and reliability of the model in real-world applications.

"Enhanced Detection of Suspicious Financial Transactions Using Autoencoder and Risk-Based Methods", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 7, page no.b668-b671, July-2025, Available :http://www.jetir.org/papers/JETIR2507176.pdf

"Enhanced Detection of Suspicious Financial Transactions Using Autoencoder and Risk-Based Methods", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 7, page no. ppb668-b671, July-2025, Available at : http://www.jetir.org/papers/JETIR2507176.pdf