A Hybrid Data Resampling Algorithm Combining Leader and SMOTE for Classifying the High Imbalanced Datasets

Objective: The traditional classifiers are ineffective in classifying the imbalanced datasets. Most popular approach resolving this problem is through data re-sampling. A hybrid resampling method proposed paper that reduces misclassification all classes. Method: employs Leader algorithm for under sampling and SMOTE oversampling. It generates desired number of samples both classes based on overcomes over-fitting under-fitting issues. Findings: To evaluate performance work, it tested 13 high datasets obtained from keel repository results compared with state-of-the-art methods such as SMOTE+Tomek Links, SMOTE+ENN, SMOTE+RSB*. From experiment observed among datasets, outperforms 12 produces same result 1 dataset. rates minority majority more suitable extreme Novelty: This research work introduces a novel classification by combining machine learning algorithms domain-specific knowledge resulting significantly improved accuracy to methods. uniqueness utilization required ratio instead balancing improves data. Keywords: Imbalanced Data; Leader; SMOTE; Hybrid Sampling; Resampling; Classification