A Weighted Fuzzy Clustering Algorithm Based on Density

The selection of the initial points influences greatly on the results of traditional partition clustering algorithms. If the algorithm starts with improper points, it will return a local minimal value. A novel weighted Fuzzy C-means algorithm (shorted by DFCM) was proposed to overcome the shortcoming. Its accuracy and effect are improved through the calculation of the relative density differences attributes, using the results of the center to determine the initial method for clustering. The numerical experiments proved that the DFCM algorithm not only can obtain better results steadily but also can distinguish the importance of each attributes.