MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202621078203 A) filed by Ram Kumar; and Monica Sankat on June 24, 2026, for Self-Discovering Counterfactual Federated Learning System For Consumer Behaviour Prediction.
Inventors include Ram Kumar; Monica Sankat; Dr Abdul Rashid; Dr. Prasad Begde; Dr. Pankaj Kumar Jha; and Dr. Himadri Srivastava.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: ABSTRACT SELF-DISCOVERING COUNTERFACTUAL FEDERATED LEARNING SYSTEM FOR CONSUMER BEHAVIOUR PREDICTION A system (100) for self-discovering counterfactual federated learning for consumer behaviour prediction comprises a plurality of user devices (110) and a central server (140). Each user device (110) comprises a data collection unit (112) configured to record user interaction signals locally without transmitting raw user interaction data to the central server (140), a causal drift engine (116) configured to apply a structural causal model (114) to perform a causal intervention producing a counterfactual behavioural distribution and to compute a genuine-change score and an event-change score, an uncertainty propagation engine (118), an adaptive gating model (122) producing three continuous weights for a local prediction model (124), an event-adapted model, and a shared base model (158), and a dynamic adapter generator (128). The central server (140) comprises a trustworthiness module (142), an event discovery engine (144), and a confidence-weighted selective aggregation engine (156). (Fig. 1) Dated This 24th day of June 2026 (Kuldeep Singh) Authorized Agent for the Applicant, Indian Patent Agent Regn No. IN/PA-4358
Disclaimer: Curated by HT Syndication.
