Abstract:

By introducing the idea of lower and upper approximations, rough K-means has become a powerful algorithm for clustering analysis with overlapping clusters. Its derivative algorithms such as rough fuzzy K-means and fuzzy rough K-means describe the uncertain objects located in the boundaries in detail, thus improving the clustering effect. However, these algorithms do not fully consider the influence of the factors, such as the distance between the data centers of the clusters and the average center and the density of the data distributed in the neighborhood, on the clustering accuracy. Aiming at this problem, a local density adaptive measure method is proposed to describe the spatial characteristics of data objects in a cluster. A rough K-means clustering algorithm based on local density adaptive measure is given. Comparative experimental results of real world data from UCI demonstrate the validity of the proposed algorithm.

Key words: rough clustering, K-means, local density measure, rough sets