A New Oversampling Technique Based on the Contribution of Regional Information

Class imbalance is one of the key factors affecting the performance of classifiers. SMOTE, as a widely used oversampling method, generates new samples through linear interpolation but does not fully consider the local characteristics of sample distributions. To address this issue, this paper proposes a region-based information contribution oversampling method (RIC-SMOTE). The method first analyzes the distribution characteristics of minority class samples, divides them into different regions, and then dynamically determines the number of samples to be generated based on the information contribution of each region. Experimental results demonstrate that RIC-SMOTE can effectively improve classifier performance.