An Improved KNN Classification Method Based on Variable Precision Rough Set

KNN classifier is a simple, non-parametric, high efficiency algorithm. But it has the defect that the classification efficiency will increase as the enlargement of data scale. This paper put forward a new KNN classification method based on rough set on the research of algorithms presented by Yu Ying and Heping Gou. The innovation of this algorithm is that the kernel, negative domain and boundary region of variable precision rough set are used to be the indexes of measuring the intra-class, extra-class and the class boundary of training sample set. The samples of intra-class, extra-class and the class boundary that are to be classed have differential treatment when judging the category. In this way, the scale of training sample set is effectively reduced and the efficiency and precision of classification are improved. At last, the category function is improved to reflect the category of samples to be classed. The experimental results show that the method can obviously improve the classification efficiency and accuracy.