MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641126336 A) filed by Prof. Pamela Bera; Prof. Bhavya K; Prof. Sonam Kumar; Supriya L; Veditha Yadav D; Vishnu R N; and Ashwin U on October 23, 2026, for Safeguarded Threshold Selection For An Integrated Money-Mule Detection System.
Inventors include Prof. Pamela Bera; Prof. Bhavya K; Prof. Sonam Kumar; Supriya L; Veditha Yadav D; Vishnu R N; and Ashwin U.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: ABSTRACT OF THE INVENTION The present invention relates to a computer-implemented system and method for reliably determining a decision threshold that converts a machine-learning-generated transaction risk score into a suspicious or non-suspicious classification, for use within a money-mule detection and investigation platform. A hybrid detection module comprising a supervised classification model and an unsupervised anomaly-detection model generates a unified risk score for each transaction. A threshold-selection module determines a decision threshold against a held-out validation partition, incorporating safeguard procedures that select a predetermined default threshold when the validation part is degenerate, resolve threshold ties toward a stricter outcome, and generate an alert when an abnormal proportion of transactions is classified as suspicious, including when a previously selected threshold is reused against newly received data. The system further comprises a transaction-network analysis module identifying structural fraud patterns and an explainability module generating feature-level risk-score attributions, and supports transaction data that is fully labelled, unlabelled, or a mixture of both.
Disclaimer: Curated by HT Syndication.
