MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641049830 A) filed by Koneru Lakshmaiah Education Foundation, Guntur, Andhra Pradesh, on April 19, for 'a hybrid anomaly detection system integrating supervised and unsupervised learning for enhanced detection in dynamic workflow environments.'

Inventor(s) include Mr. Arun Kumar Bandlamudi; and Dr. Sunitha Pachala.

The application for the patent was published on May 1, under issue no. 18/2026.

According to the abstract released by the Intellectual Property India: "The present invention provides a Hybrid Anomaly Detection System (HADS) that seamlessly integrates supervised and unsupervised machine learning for superior anomaly identification in dynamic workflow environments. The system ingests real-time multi-modal data from evolving processes such as manufacturing lines, logistics, or business automation platforms. A Workflow Change Detector identifies concept drift using statistical tests, triggering adaptive re-weighting. Supervised models (e.g., XGBoost) leverage labeled historical anomalies for precision on known patterns, while unsupervised ensembles (Isolation Forest + Autoencoders) detect novel outliers without labels. A novel Hybrid Fusion Layer computes a weighted anomaly score, with weights dynamically optimized by a reinforcement learning agent that learns from expert feedback and operational outcomes. This enables the system to adapt continuously to workflow changes, minimizing false positives and negatives. An explainable alert generator provides prioritized notifications with feature attributions. The invention overcomes limitations of static or single-paradigm systems by offering real-time adaptability, scalability, and continuous improvement. It is implemented via standard computing hardware/software and significantly enhances detection reliability in non-stationary environments. Practical case studies demonstrate 15-30% performance gains. The system is robust, interpretable, and industrially deployable."

Disclaimer: Curated by HT Syndication.