A Hybrid Weighted Nearest Neighbor Approach to Mine Imbalanced Data

Classification of imbalanced data has drawn significant attention from research community in last decade. As the distribution of data into various classes affects the performances of traditional classifiers, the imbalanced data needs special treatment. Modification in learning approaches is one of the solutions to deal with such cases. In this paper a hybrid nearest neighbor learning approach is proposed for mining binary imbalanced datasets. This hybridization is based on different K for different classes as large K for large classes and small K for small classes and with weighted concept as small weights for large classes and large weight for small classes. The merger of dynamic K improves the performance of weighted approach.