Multi-objective Evolutionary Clustering Algorithm And Its Application

As an effective method for pattern recognition,clustering has been widely applied to many areas in real world.With new techniques and new ideas constantly emerging,the approaches to clustering are being more and more diverse,resulting in space for improvement.Under this situation,based on the evolutionary algorithm which has a superior performance for optimization problems,in this paper,we have made some research work about clustering and its application in image segmentation,mainly including following two parts:(1)In order to tackle the clustering problem with an unknown number of clusters,this paper has proposed a novel differential evolution approach to multi-objective clustering.Determining the number of clusters is a basic yet challenging problem in clustering analysis.In one hand,the optimal number of clusters varies according to different evaluation criteria,user preferences or demands,hence it makes sense to provide the user with multiple clustering results for different number of clusters.On the other hand,increasing the number of clusters usually optimizes the within-cluster compactness while deteriorates the between-cluster separation,therefore,selecting an appropriate number of clusters is,in fact,a multi-objective optimization problem which needs to choose a balanced solution among a set of tradeoffs between the minimum number of clusters and the maximum compactness or separation of clusters.As a result,we directly take the number of clusters as one optimization objective and optimize it with another clustering validity index simultaneously by a newly designed multiobjective differential evolution algorithm.The proposed algorithm obtains a nearly Pareto-optimal set containing multiple clustering results for distinct number of clusters,in a single run.Besides,the appropriate clustering solution can be flexibly selected from the solution set as the final clustering result.Comparative experiments on several datasets have verified the practicability and effectiveness of our proposed algorithm.(2)Performance of segmenting noisy image by fuzzy clustering can be easily affected by the noises and imaging artifacts.In fuzzy clustering,it generally is a key point to control the level of fuzziness which determines the degree of overlap between different clusters.Hence,we have proposed a multi-objective evolutionary fuzzy clustering with adaptive level of fuzziness for the challenging noisy image segmentation.By optimizing the entropy of membership values and the fuzzy compactness within clusters concurrently,the proposed method can generate multiple compact fuzzy clusters under different levels of fuzziness in a single run without any level of fuzziness specified.Moreover,it automatically evolves some crucial control parameters through encoding them into chromosomes,and it is these parameters that balance the image detail preservation and noise insensitiveness.Experimental results on several synthetic and real-world images demonstrate that the proposed method can improve the segmentation quality and is more robust to noises.