MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641104372 A) filed by Vetrivel M; and Dr. Arul Joseph G on August 31, 2026, for System And Method For Detecting Fraudulent Financial Transactions Using Fractional-Order Feature Mapping And Geometric Prototype- Based Risk Classification.
Inventors include Vetrivel M; and Dr. Arul Joseph G.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: Financial fraud detection requires classifiers that adapt to evolving fraud patterns while operating within the latency constraints of transaction authorization. This invention provides a computer-implemented system and method. referred to as the Fractional Geometric Risk Neural Netw_o r-k (FO—R-N —et), for classifying financial transactions according to fraud risk. Raw transaction feature data is transformed using a fractional-order mapping that captures memory and long-range dependence not present in integer-order feature representations. The fractionally-mapped features are evaluated against a set of learned geometric prototypes representing characteristic legitimate transaction behaviour, yielding a prototype distance score, and against a local neighbourhood-density-based abnormality score sensitive to novel patterns not resembling any existing prototype. The two indicators are combined by an adaptive fractional risk function whose fractional-order parameter varies according to local feature density, rather than applying a fixed decision rule uniformly across the feature space. The resulting risk score classifies the transaction as fraudulent or legitimate. The system operates on standard transaction feature data without requiring construction of a relational transaction graph, and the underlying fractional geometric mapping and risk-scoring architecture is adaptable, upon retraining, to other structured-data anomaly classification domains.
Disclaimer: Curated by HT Syndication.
