MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202621052116 A) filed by Symbiosis International Deemed University on April 23, 2026, for Adaptive Relational Synthetic Minority Oversampling System And Method For Imbalanced Data Classification In Machine Learning.
Inventors include Dr. Snehlata Wankhade; Anushri Adapawar; Anushka Poshattiwar; and Tanvi Bandebuche.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT ADAPTIVE RELATIONAL SYNTHETIC MINORITY OVERSAMPLING SYSTEM AND METHOD FOR IMBALANCED DATA CLASSIFICATION IN MACHINE LEARNING The present invention provides an adaptive relational synthetic minority oversampling system (100) and method for addressing class imbalance problems in machine learning classification tasks. The system comprises a data acquisition module (110), preprocessing module (120), stratified splitting module (130), imbalance ratio computation module (140), adaptive balancing module (150), relational SMOTE module (160), classifier training module (170), and evaluation module (180). The imbalance ratio computation module (140) automatically categorizes datasets as balanced, moderately imbalanced, or highly imbalanced based on computed ratios. The adaptive balancing module (150) selects appropriate resampling strategies accordingly. The relational SMOTE module (160) generates synthetic minority samples using nearest neighbor interpolation with controlled Gaussian noise injection according to the formula: Synthetic = x_i + r * (x_n - x_i) + noise. Experimental results demonstrate significant improvements in minority class detection metrics including precision, recall, and F1-score while maintaining robust overall classification performance. The invention is applicable across diverse domains including healthcare, finance, cybersecurity, and manufacturing. [
Disclaimer: Curated by HT Syndication.
