The Research On Clustering Analysis And Its Application In Data Mining Of Mobile Communication Enterprise

Data mining is the procedure of extracting of implicit, original, useful knowledge in the database, which is widely applied in many fields in recent years. Clustering analysis is one of the main technology measures in the research on data mining with a mass of theories and methods achieved. With the establishment of Decision Support System such as data warehouse and the requirement of business intelligence in the data intensive enterprises, data mining has been used in many new applications and the research on data clustering is faced with a lot of new challenges. The mobile communication enterprise is one of the typical data intensive enterprises. With the aggravation of the telecommunication market competition, subdividing the customers and supplying different marketing and service to different customers group has already become instant demand to the telecommunication enterprises at the present time. A new algorithm clustering the data sets with mixed numerical and categorical values is researched based on the requirement of customer segmentation and the characteristic of the data in the mobile communication enterprise and the marketing strategies is proposed in the paper. The research is summarized as follows:1. The technology of data mining is introduced briefly, and clustering analysis in data mining is disserted, involving the methods, characteristics and sorts of clustering,with emphasis on clustering the data sets with mixed numerical and categorical values. Then Fuzzy K-Prototypes clustering algorithm is introduced with its advantages and disadvantages.2. In order to overcome the defects in the Fuzzy K-Prototypes clustering algorithm including sensitivity to the initial data and being easy to run into the local optimization, A new hybrid clustering algorithm based on PSO (particle swarm optimization) and Fuzzy K-Prototypes algorithm is proposed, by using PSO to determine the centroids of clusters and taking the clustering result of PSO as the initialized value of the Fuzzy K-Prototypes. The results show that the proposed algorithm is superior to FKP algorithm with a better astringency and stability. In addition, considering the particular contributions of different features,we use ReliefF algorithm to assign the weights for every feature.3. The application of clustering technology in telegraphic customer segmentation is researched. The article studies the basic theory, methods and process of customer segmentation. After investigating the original data of the customers from the Hunan Mobile Co.Ltd's BASS (Business Analysis Support System) platform, we do the customer segmentation based on customer's behavior trait and consumption psychology with clustering technology, and use the customer segmentation result to help making marketing strategies.