This paper mainly improves the choice of the initial clustering center and isolated point problem of K-means algorithm. Firstly, we calculate the distances between all the data objects, and eliminate the influence of isolated points according to the ideas of the distance sum. Then we put forward a new method of the initial clustering center selection. In experiment part, we make the comparison between the improved algorithm and the original algorithm through the experiment. Experiment results show that our improved algorithm obviously decreases the influence of the isolated point, and the clustering result is closer to the actual data distribution.
