MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641086809 A) filed by M R Archana Jenis; Dr. Dharavathu Radha; Dr. Gali Avinash Kumar; P. V. Samuel Blessed Nayagam; D. V. Viji Nirmala; Akhila Judit Nisha S; Dr. N. Ramya; T. Saroja; Kalpita Dutta; and Boggula Roja Ramani on July 15, 2026, for Quantum-Enhanced Attrition Prediction And Retention Intelligence System For Students And Employees.
Inventors include M R Archana Jenis; Dr. Dharavathu Radha; Dr. Gali Avinash Kumar; P. V. Samuel Blessed Nayagam; D. V. Viji Nirmala; Akhila Judit Nisha S; Dr. N. Ramya; T. Saroja; Kalpita Dutta; and Boggula Roja Ramani.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: The proposed invention, titled "Quantum-Enhanced Attrition Prediction and Retention Intelligence System for Students and Employees," relates to an intelligent and automated system that utilizes Federated Learning combined with a multimodal quantum-classical classifier to predict attrition risk while preserving the privacy of sensitive student and employee data. Attrition in educational and organizational settings, being a critical economic and operational concern, is difficult to detect through conventional models that rely solely on structured performance data, resulting in delayed intervention and preventable loss of students or employees. Traditional predictive approaches, such as centralized machine learning models trained on tabular metrics alone, are context-blind, inconsistent in identifying underlying behavioural risk, and require aggregating sensitive records in a manner that conflicts with data sovereignty requirements. The invention addresses these challenges by introducing a decentralized, privacy-preserving framework that performs accurate, verifiable, and behaviourally aware attrition prediction without centralizing raw data. The system enables each participating institution or organization to locally train a Variational Quantum Classifier operating on a fixed-qubit circuit using its own private data, transmitting only optimized model weights, rather than raw records, to a central server through a Federated Averaging protocol. The resulting local models are further enhanced through a multimodal fusion layer that combines structured tabular metrics with unstructured acoustic tone and textual sentiment indicators, encoding the fused features into quantum states via angle encoding for processing through a strongly entangling variational circuit. A heuristic risk-calibration mechanism dynamically increases predicted risk scores upon detection of specific behavioural triggers, while an integrated explainability module identifies the most influential risk factors underlying each prediction in real time. The invention also integrates an administrative console that supports batch processing, real-time inference, and interpretable risk dashboards. By automating privacy-preserving, multimodal attrition prediction, the proposed system eliminates the risks of centralized data exposure and context-blind forecasting, enabling consistent, sensitive, and explainable identification of at-risk individuals. It can be deployed in educational institutions, corporate human resources platforms, and other organizational retention-management systems.
Disclaimer: Curated by HT Syndication.
