A Graphic Clustering Algorithm Based on MMAS

An adaptive graphic clustering algorithm (AGCM) based on MAX-MIN Ant System (MMAS) is proposed. A "similarity" between objects in the space of object attributes is defined, and "similarity" weights on the directed edges of a pheromone map are assigned. The weight of the similarity on every edge is adaptively updated by the pheromone left by ants in seeking process. Pheromone trail updating adopts self-adaptive strategy. In contrast to the usual ant colony clustering algorithms, this paper maps the pheromone values into the interval [0,1]-Because of this mapping transformation, the scope of parameter epsiv can\´t be changed too much, and some principles can be followed. In this algorithm neither the number of data clusters nor the initial guessing of cluster centers is required. Experimental results demonstrate this algorithm is superior to the existing ant colony clustering algorithms (LF and A³ CD) with shorter running times and better qualities.