Improved K-means Clustering Algorithm based on Genetic Algorithm

Abstract

Through comparison and analysis of clustering algorithms, this paper presents an improved K-means clustering algorithm. Using genetic algorithm to select the initial cluster centers, using Z-score to standardize data, and take a new method to evaluate cluster centers, all this reduce the affect of isolated points, and improve the accuracy of clustering. Experiments show that the algorithm to find the initial cluster centers is the same location, objective function value is smaller, the clustering effect is better and more stable when it has the outlier data, and it applies not only to simple data sets, but also to more complicated data sets.

DOI : http://dx.doi.org/10.11591/telkomnika.v12i3.4262