A Modified K-Means Algorithm - Two-Layer K-Means Algorithm

In this paper, a modified K-means algorithm is proposed to categorize a set of data. K-means algorithm is a simple and easy clustering method which can efficiently classify a large number of continuous numerical data of high-dimensions. Moreover, the data in each cluster are similar to one another. However, it is vulnerable to outliers and noisy data, and it spends much executive time in classifying data too. Noisy data, outliers, and the data with quite different values in one cluster may reduce the accuracy rate of data matching obtained by a pattern matching system since the cluster center cannot precisely describe the data in the cluster. Hence, this study provides a two-layer K-means algorithm to solve above problems. In experiment, several well-known data sets are used to evaluate the performance of proposed algorithm, and the two-layer K-means algorithm can give expressive experimental results.