Research Of Clustering Algorithm Based On Quantum-Behaved Particle Swarm Optimization Algorithm

Unsupervised clustering analysis, regarded as a very important technology of data mining, has been applied to pattern recognition, image processing, biological computing and so on. This paper studies the improved Quantum-behaved Particle Swarm Optimization(QPSO) Algorithm, and uses it to optimize the K-means algorithm, the FCM algorithm and the KFCM algorithm.Although the QPSO algorithm's global searching ability is better than that of the Particle Swarm Optimization(PSO) algorithm's, it tend to be premature like other algorithms. In order to reduce the happening of this kind of situation, this paper will introduce an improved QPSO algorithm. Simulation results on 4 benchmark functions show that the algorithm is superior to the PSO and QPSO.The K-means, the FCM algorithm and the KFCM algorithm have their own advantages: the K-means algorithm's ideas are easily understandable and easy to be operatde; the FCM algorithm has the deep mathematical foundation and the fast computing speed; the KFCM algorithm is good at dealing with nonlinear clustering. But they have three common shortcomings: high requirements for initial value, sensitive to outlier, not a globally convergent algorithm. The improved QPSO algorithm has the stronger capability of global search, the lower demands at initial value in looking for optimum. So the K-means, the FCM algorithm and the KFCM algorithm were improved by the improved QPSO algorithm. The experimental data show that these method is available and efficient.In this paper, the designed algorithm got the implementation both in the Matlab environment.