Weighted kNNModel-Nased Data Reduction and Classification

A weighted kNNModel-based data reduction and classification algorithm, called wkNNModel, is proposed in this paper which aims to find some more meaningful representatives to replace the original dataset for further classification. Each representative is formed by an instance and its weighted neighbourhood satisfies a predefined threshold. Compared to kNNModel the proposed method, as an alterative to other kNN algorithms, further alleviates the effect of abnormal data, i.e. noisy or boundary disturbance data, to data reduction and classification, thus contributing to the improvement of data reduction rate.