MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641087045 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on July 16, 2026, for A Hybrid Machine Learning System For Financial Fraud Detection Using Ctgan-Based Synthetic Data Generation And Ensemble Classification.
Inventors include Dr. Vasavi Ravuri; Dr. R. Vijayasaraswathi; K. Swathi; M. Mohana Deepthi; M. Suraj; and K. Manish.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: As digital transactions are growing rapidly, financial fraud has emerged as a serious issue that needs intelligent and scalable solutions. Conventional money laundering detection systems tend to have difficulty with class imbalance, scarce labelled data, and poor generalisation ability. This work introduces a hybrid machine learning framework combining synthetic data generation using CTGAN Synthesizer and ensemble-based fraud detection models to improve the detection of fraudulent financial transactions. The system proposed uses generative models to generate high-quality synthetic fraud data, which is blended with actual transaction data to train machine learning algorithms efficiently. Sophisticated preprocessing, feature engineering including behavioural and temporal patterns boosts the model's capacity to learn sophisticated fraud indicators. The architecture includes CTGAN Synthesizer and Random Forest classifier, and measures model performance using industry-standard metrics like ROC-AUC and precision-recall analysis. Experimental results show a high accuracy of 99%, precision and recall of 0.97 for fraud classes, and a ROC-AUC score of 0.9816, indicating the model’s strong generalisation ability and a significant reduction in false positives. The method effectively addresses the challenges of data sparsity and class imbalance, proving its utility in real-world financial fraud detection scenarios.
Disclaimer: Curated by HT Syndication.
