Abstract:

Aiming at the problem that affinity propagation algorithm has a difficult to deal with highdimensional data, the authors put forward an selfadaptive affinity propagation algorithm based on singular value decomposition. With the aid of singular value decomposition introdued, the highdimensional data were reconstructed and dimensions were reduced to eliminate the redundant information, based on which, a nonlinear function strategy was adopted to adjust the damping factor adaptively and improve the clustering performance of the algorithm. Experimental results show that the proposed algorithm has obviously better clustering performance than the traditional algorithm on clustering accuracy and the number of iterations.


Key words: affinity propagation clustering algorithm, singular value decomposition, nonlinear function strategy, damping factor