MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641096189 A) filed by Dr. S. Saraswathi; Dr. M. Suresh; Dr. C. Hariharasudhan; Dr. T. Sudha; Dr. M. Manimekala; Dr. D. Suryaprabha; Ms. M. Varsha; Ms. U. Ramya; Ms. J. Jobi Jesmitha; and Dr. R. A. Ayyapparajan on August 09, 2026, for System And Method For Automated Detection Of Data Drift And Adaptive Machine Learning Model Retraining Using Multi-Dimensional Degradation Scoring.
Inventors include Dr. S. Saraswathi; Dr. M. Suresh; Dr. C. Hariharasudhan; Dr. T. Sudha; Dr. M. Manimekala; Dr. D. Suryaprabha; Ms. M. Varsha; Ms. U. Ramya; Ms. J. Jobi Jesmitha; and Dr. R. A. Ayyapparajan.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: ABSTRACT OF THE INVENTION The present invention relates to a computer-implemented system and method for automated detection of machine learning model degradation and adaptive model retraining. The system continuously monitors deployed machine learning models by evaluating data drift, concept drift, feature-level anomalies, and model performance degradation using a Multi-Dimensional Degradation Scoring Engine (MDSE). The computed degradation indicators are integrated into a composite degradation score through a dynamic context-aware weighting mechanism. An adaptive threshold module classifies the operational state of the model and a decision engine automatically determines an appropriate response, including monitoring, alert generation, incremental retraining, or full model retraining. A retraining pipeline validates the updated model before deployment and recalibrates monitoring parameters following successful deployment. The invention provides an automated, scalable, and model-agnostic framework for maintaining the accuracy, reliability, and operational performance of deployed machine learning systems.
Disclaimer: Curated by HT Syndication.
