Research On Fuzzy Clustering Analysis And Its Validity

Cluster analysis is an important branch of unsupervised pattern recognition.Because of having established the sample that describes the uncertainty of categories,and being able to reflect the real world more objectively,fuzzy clustering has become the mainstream of cluster analysis study.Fuzzy clustering has been widely used in pattern recognition,data mining,image processing and many other fields.The contribution of this paper exists in two aspects:Since the traditional kernel fuzzy clustering algorithm does not take into account of the respective contribution made by different features,and it also has a shortcoming of easily falling into a situation of a local optimum,an improved kernel fuzzy clustering algorithm is put forward. Combing the advantage of the global optimum that the genetic algorithm the improved kernel fuzzy clustering algorithm constructs a simple and effective fitness function that can avoid plunging local optimum.This improved algorithm gives every feature a weighted coefficient,in which the ReliefF algorithm is used to assign the weights for every feature.Comparing with the traditional algorithm, this one has made some significant progresses,and the experimental result has proved its effectiveness.A new cluster validity index is proposed that determines the optimal partition and optimal number of clusters for fuzzy partitions obtained from the fuzzy c-means algorithm.The proposed validity index combines compactness measure and separation measure between clusters.The compactness measure is obtained by computing inter-cluster weighted the square of error.The separation measure is obtained by computing similarity between fuzzy clusters.In the three synthetical datasets and three well-known real datasets the experiment proved that the proposed index is superior effectively in comparison to other indexes.Full text is divided into five chapters,each chapter is as follows: Chapter 1 is the introduction that gives the research background and significance,briefly presenting the main scope of this research and its structural arrangements.Chapter 2 is the literature review that looks back at the status quo of the cluster analysis and fuzzy clustering algorithm research,and related theories.Chapter 3 focuses on an improved Kernel-based fuzzy clustering algorithm.It also conducts a comparative experiment and analysis on the clustering effect of the algorithm.Chapter 4 proposes a new validity indicator of fuzzy clustering.A comparative study of the new indicator and existing indicator is also conducted in the six data sets.Chapter 5 is the concluding part of this thesis.It presents a summary of the current research and its prospect.