Clustering algorithm based on Delaunay triangulation density metric

According to the problem that K-Means clustering algorithm fails to correctly distinguish non-convex shape clusters, computation mode of distance in the algorithm is changed and density metric mode which can reflect the characteristics of data themselves is adopted instead. In the mode, Delaunay triangulation graph which has the advantages of nearest neighbour and adjacency is introduced to compute density. Simultaneously, because K-Means clustering algorithm itself belongs to a global optimization problem, chaos optimization dedicated to global optimization is applied to optimize objective function for the sake of obtaining global minimum solution. Experiments results indicate that clustering algorithm based on Delaunay triangulation density metric can find arbitrary non-convex shape clusters.