MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202621047943 A) filed by Moresh Madhukar Mukhedkar; Dhairyasen Deshmukh; Avneet Singh Arora; Onkar Jadhav; Zaid Shaikh; Vivek Patil; Shakil Tamboli; and Shweta Koparde on April 15, 2026, for Intelligent Academic Stream Recommendation System Using Ml.
Inventors include Moresh Madhukar Mukhedkar; Dhairyasen Deshmukh; Avneet Singh Arora; Onkar Jadhav; Zaid Shaikh; Vivek Patil; Shakil Tamboli; and Shweta Koparde.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: Choosing the right academic stream after completing secondary education is one of the most consequential decisions a student can make, with lasting implications on their career trajectory and personal fulfilment. Conventional approaches to academic counselling rely heavily on the subjective opinions of parents, teachers, and peers, frequently resulting in misalignment between a student's natural aptitudes and the stream they eventually pursue. This paper presents the design, development, and empirical evaluation of an Intelligent Academic Stream Recommendation System (IASRS) that leverages supervised machine learning classification algorithms to provide objective, data-driven guidance. The proposed system collects structured information on a student's academic grades across core subjects, scores on standardized aptitude assessments, and self-reported interest levels across distinct knowledge domains. Three widely adopted classification algorithms—Decision Tree (DT), K-Nearest Neighbours (KNN), and Random Forest (RF)— were implemented, trained, and rigorously evaluated on a curated student dataset. Experimental results demonstrate that the Random Forest classifier consistently outperforms the other models, achieving an overall accuracy of 94.2% on the held-out test set. The system outputs a ranked recommendation of academic streams—Science, Commerce, or Arts and Humanities—together with confidence scores, thereby equipping students and educational counsellors with transparent, interpretable insights. Ablation experiments further reveal that subject-specific grades and aptitude scores collectively constitute the most discriminative feature subset for stream prediction. The proposed framework is computationally lightweight, requiring no dedicated hardware beyond a standard personal computer, and can be readily integrated into existing educational information systems.
Disclaimer: Curated by HT Syndication.
