An Improved Semi-supervised K-Means Algorithm Based on Information Gain

The traditional K-means algorithm is sensitive to the initial center, and equates the importance of dimension data for multidimensional data. So it is unable to block the effects of dimensional data dimension, nor can it well reflect the influence of each dimension of clustering. The semi-supervised clustering introduces a small amount of sample points, so that it can significantly reduce the number of iterations, as well as increase the efficiency of clustering accuracy and iteration. This paper introduces ideas of the information gain weighted to the semi-supervised K-means algorithm. By using a small amount of marked samples to the information gain weight calculation and determination of the initial center, the algorithm in this paper obtains the clustering effect of higher quality, and maintains the stability of the cluster.