The Fuzzy Classification Evaluation Research Based On Clustering Validity Index

In recent years, the theory and empirical research of comprehensive evaluation technology have the very big development, and various evaluation methods get rid of the stale and bring forth the fresh. In fact, in the multi-index comprehensive evaluation, the index of many evaluation objects itself is fuzzy. As an integral part of comprehensive evaluation method, the fuzzy comprehensive evaluation method plays an irreplaceable role increasingly in the multi-index comprehensive evaluation. In fact, the fuzzy comprehensive evaluation can be divided into fuzzy sorting evaluation and fuzzy classification evaluation, and fuzzy classification evaluation mainly refers to two categories, fuzzy clustering and pattern recognition.With the advent of the big data age, cluster analysis has received the widespread attention in the practical application as an effective tool in data mining technology. Since clustering is an unsupervised classification process, it has no priori information of data set. It is necessary to test and evaluate the validity of clustering results. Through the cluster validity analysis, we can not only judge the validity of clustering results, but also get the best clustering results.In this paper, I make an in-depth study of the fuzzy clustering validity. And on the basis of the existing fuzzy clustering validity index, I redefine a new compactness and separation degree, and build a new fuzzy clustering validity index with the ratio of compactness and separation degree. In this paper, I theoretically prove that the new validity index essentially overcomes the two defects of the validity index VXB proposed by Xie and Beni. And I carry out some simulation experiments to test the reliability of the new validity index with three groups of Gaussian random data and eight groups of real data from the UCI database. At the same time, I compare the new validity index with other five existing indices. And through the experiment, we find that the new validity index compared with other five indicators can get the optimal clustering number more accurately. Thus illustrates the advantages of the new validity index.Construction of validity index is not once and for all because the existing fuzzy clustering validity indices have their advantages and limitations. At the same time, we will face a choice problem with many complex fuzzy clustering algorithms. Therefore, this paper proposes a combination of fuzzy clustering algorithm by building a combined model with the multiple fuzzy clustering algorithm and fuzzy clustering validity index. And we make weighted summation of all the results of combination algorithm. Then we use weighted voting by machine to select the optimal clustering number with highest votes. In this paper, we simplify the model into a combination model with single fuzzy clustering algorithm and multiple fuzzy clustering validity indexes in order to test the effectiveness of the combination model more intuitively. In this paper, we focus the Fuzzy c-means clustering algorithm FCM and the six fuzzy clustering validity indices of VPC, VPE, VMPC, VXB, VK and Vnew,and make the simulation experiments to verify the validity and reliability of the fuzzy clustering combination model. Thus, in this paper, we verify the effectiveness and reliability of the combination model to a certain extent.At the same time, in order to further study the validity and reliability of the new fuzzy classification validity index and fuzzy classification combination evaluation method. I make the fuzzy classification evaluation of economic development level of11cities of Zhejing province.