Multi-objectiye Optimization Algorithm And It’s Applications

With the rapid development of computer technology, we need to face to theexponential rising trend of data, data mining algorithm by which can mining theinternal contact is becoming an increasingly important and data mining is becoming anincreasingly urgent problem to be solved. As a very important branch of data mining,Clustering Algorithm has attracted an increasingly number of researchers. In this paper,we employed multi-objective optimization on clustering algorithms and imagesegmentation, the main work of this paper is summarized as follows:(1) An Immune clonal multi-objective optimization for clustering andclassification algorithm is proposed in this dissertation. In this section, we employedthe Immune clonal multi-objective algorithm to simulate clustering and classificationframework. The non-dominated sorting based immune clone operation, immunegenetic operation and antibody updating operation in our algorithm made the algorithmobtained more non-dominated solutions and this leaded to a better effect ofclassification. Two groups of experiments are listed in this paper to test theclassification accuracy of our algorithm and the test results perfectly matched ourtheoretical analysis that our algorithm had better diversity, uniformity and convergence,and also better classification accuracy.(2) An Immune clonal multi-objective optimization algorithm for imagesegmentation is proposed in this dissertation. We use gray-level co-occurrence matrixand Gabor filter to extract image features, through the use of pre-watershed imagesegmentation processing to solve the problem of speed on the issue of the problem ofinsufficient. The experiments listed in this paper show the image segmentation hasmore uniformity and more accuracy.(3) A Multi-objective particle swarm optimization algorithm on Nystr m spectralclustering is proposed in this dissertation. By affinity functions and the uses ofmulti-objective particle swarm optimization algorithm, we can select more diversityand uniformity Nystr m representative point. Effectively improve the Nystr m spectralclustering classification accuracy. Simulation results show that the algorithm has betterstability and classification accuracy.