Research and experiment on Affinity Propagation clustering algorithm

This paper introduces Affinity Propagation (AP) clustering algorithm, which is intensively researched by some scholars owing to its advantage of fast speed and no need of setting the initial clusters manually. Mainly analyzed the characteristics of Affinity Propagation clustering algorithm at first, and then compared several principle similarity calculating methods based on Euclidean distance and Mahalanobis distance and etc. Experiment on AP clustering algorithm were done with the parts of the UCI data sets, thus the effectiveness of this algorithm was verified. Finally, the experimental results were analyzed in general.