MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641100172 A) filed by Karpagam College Of Engineering; Karpagam Academy Of Higher Education; R. Ram Prashath; and Jeffery Alwin R on August 19, 2026, for A Data-Centric Experimental Framework For Evaluating Distance-Based Dataset Reduction.
Inventors include R. Ram Prashath; and Jeffery Alwin R.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: This research tackles the growing challenge of managing massive datasets in supervised machine learning without losing predictive accuracy. As data grows, redundant and near-duplicate samples often inflate computational costs and storage needs without actually helping the model learn better. To solve this, we developed a hybrid framework that layers feature engineering with a specific "diversity classification" strategy. By testing multiple distance metrics—including Euclidean, Manhattan, and Cosine—we categorize data into 'less-diverse' and 'diverse' groups. We then use clustering-based pruning to strip away the redundant "noise" while keeping the informative core. Our experiments on benchmark tabular data prove that we can maintain, or even boost, classifier accuracy while significantly cutting down on training time. This report covers our full methodology, the experimental results, and where we plan to take this framework next.
Disclaimer: Curated by HT Syndication.
