K-means Clustering Algorithm and Its Improvement Research

Abstract

Clustering is a typical unsupervised learning method, and it is also very important in natural language processing. K-means is one of the classical algorithms in clustering. In k-means algorithm, the processing mode of abnormal data and the similarity calculation method will affect the clustering division. Aiming at the defect of K-means, this paper proposes a new similarity calculation method, that is, a similarity calculation method based on weighted and Euclidean distance. Experiments show that the new algorithm is superior to k-means algorithm in efficiency, correctness and stability.