Graph theory with Modify-edge Clustering Algorithm Based on Maximum Weighted Entropy

Combining with the graph theory clustering methods, an entropy-objective function algorithm was proposed for clustering. The edge which connects the vertices in non-orientation graph was redefined according to the distribution and distance of the data set. The objective function of the weighted entropy based on intra-variance in cluster and variance between clusters was built. The cluster result for the data set is derived from the maximum objective function. This algorithm doesn´t need the prior knowledge about the cluster number and the initialization centre