Application of a novel algorithm on clustering analysis

This paper intends to propose a novel clustering method, ant graphic clustering (AGC) algorithm. AGC combines the graphic clustering method with meliorative MAX-MIN Ant System. A "similarity" between objects in the space of object attributes is defined, and "similarity" weights on the directed edges of a pheromone map is assigned. The weight of the similarity on every edge is adaptively updated by the pheromone left by ants in seeking process. In this algorithm neither the number of data clusters nor the initial guessing of cluster centers is required. Experimental results show that this algorithm is easier to implement, faster to perform and the clustering quality is better.