Credibilistic Clustering Algorithms And Its Application In The Socio-Economic Problem

Birds of a feather flock together. As an important branch of pattern recognition, clustering offers the way for human to understand the world, distinguish between different things,and realize the similarity between different things. Researchers have proposed various kinds of clustering algorithms based on different theories and methods. Fuzzy clustering algorithm has the advantage in the definition of the degree of membership for data point than other algorithms, and it has been widely applied to the stock prediction, medical research, marketing and other fields.Based on the analysis of the existing well-known fuzzy clustering algorithms, the credibility thoery is used for clustering based on the mathematical properties of credibility measure, which results in a new clustering method, called credibilistic clustering. Associated with the alternating cluster estimation method, the general framework of credibilistic clustering algorithms is given. Finally, the credibilistic clustering algorithm is applied to the study of socio-economic problem.The main contents and results of this article are as follows:1. Credibility clustering model and its algorithm. A new clustering method called credibilistic clustering is developed by introducing the credibility measure into the field of cluster analysis in this paper. To tackle the proposed model with optimal partitions, we suggest the general framework of the credibilistic clustering algorithm by integrating alternating cluster estimation into the clustering process. Besides, two special credibilistic clustering algorithms are given for the further study of this new algorithms.2. Analysis of the validity of the credibilistic clustering algorithms. In order to show the validity of the new algorithm, experiments from different aspects are processed by a series of data sets, including coincident clusters problem, noisy environment, and overall accuracy. Results show that credibilistic clustering algorithms give better classifications than FCM and PCA algorithms.3. Application of credibilistic clustering in the socio-economic problem. In order to show the validity of the credibilistic clustering algorithm in solving the social problem, credibilistic clustering algorithms are applied to classify 31 provinces into 3 clusters compared with K-means and FCM algorithms, and results show the validity of the credibilistic clustering algorithms. Besides, the classification results contribute to the construction of policy for different provinces.In this paper, the research and analysis on the credibilistic clustering algorithm not only extends the theoretical research on fuzzy clustering algorithm, and also presents the application value. In the future research, credibilistic clustering algorithms can be applied to the field of text mining and image segmentation.