On Clustering Algorithm Based On Complex Attributes Similarity And Its Applications

As a suitable technique, clustering analysis can be widely used to exactly classify data to appropriate class in market segmentation. As the corporation keep more operation and pay more attention to client, the attributes describing the client character and the client-information data increase incessantly. The high-dimension characteristic of data cause a problem called"dimension disaster"to clustering algorithm.High-dimension clustering algorithms based on selecting-dimension and decreasing-dimension have to lose much information in some dimensions. When the information of each dimension is kept, how to cluster the data is the key problem of high-dimension clustering research. This dissertation mainly discusses the clustering algorithm based on complex attributes similarity. The main work is listed as follows.(1) Firstly, the common process of clustering analysis is discussed. There are five traditional clustering algorithms and three kinds of high-dimension clustering algorithms. Two key issues of the clustering algorithm based on objects similarities, the measure function of similarity and the segmentation algorithm of graph, are researched.(2) Aiming at the characteristic of market research data, a new clustering algorithm for complex attributes is proposed based on an idea of feature similarity measurement. Firstly, the objects similarities are measured by complex attributes'distribution similarity function. Then, a graph model is constructed based on the similarity. Finally, the graph is divided and clustered. The proposed algorithm can keep the integrated data. At the same time, the algorithm can process high-dimension data and complex attributes effectively. When the parameters in the algorithm are modified, it is unnecessary to review the original date, which can improve the efficiency. The initial clustering center is unnecessary, which can reduce the effect of other conditions. Based on stochastic dataset (Random), the experimental results show that the algorithm is effectively and can find high- quality clustering.(3) The application of clustering analysis technique in market research data analysis is shown. In the application, the proposed clustering algorithm based on complex attributes similarity is used. The collected data is clustered to get the detailed result of the college telecommunication market, which can be used as the reference for the market competition of the college telecommunication market. In the application, the effectiveness of the algorithm is also evaluated.