An Optimized K-Means Algorithm Based on FSTVM

Abstract

Aiming at the text document similarity and initial center point problems of K-means algorithm, a new model is proposed, in which used a new method of Frequency-Sorted Term Vector Model(FSTVM) to represent a document, and for reducing the dimension of a document designed an automatic method to filter commonly used words, and redesigned the similarity calculation formula and optimization algorithm for the initial center. Compared with the traditional algorithms, the new proposed algorithm can get initial centers with higher quality and steadier cluster results, Experimental results prove this clustering algorithm is simple and effective.