Improved Fuzzy Clustering Algorithm And Its Application In Telecom Arrears Data Mining

With the rapid development of the telecommunications market, China's communications equipment coverage has reached a very high level, and the number of mobile phone users increased year by year. To maintain the efficient operation of the market, telecom business must divide the different fee based on different regions and different groups of people, at the same time they also should take operating strategy to prevent users from reaping huge number of arrears and buying a new SIM card. Data mining in the telecommunications field has a more broad application space with its advantages in massive information processing. The clustering technique of data mining can deal with the object clustering analysis, as the application in market analysis, it can also distinguish between different users, and depict different client group characteristics. This paper describes the characteristics of the telecommunications delinquent customers and proposes clustering algorithms which fit for the telecommunications industry data features. The analysis of the characteristics of telecommunications users in arrears provides a theoretical basis for telecom operators to develop appropriate measures.The main research contents in this paper are as follows:Firstly, we study the fuzzy C-means (FCM) clustering algorithm and then propose a new FCM clustering algorithm which based on information entropy. The new algorithm uses information entropy to initialize cluster centers and then determines the number of cluster center. This step reduces the initial error caused by clustering algorithm. Meanwhile, in order to make algorithm fit for arbitrary shape data sets, this paper quoted the merger idea, which splits into clusters of arbitrary shape categories, and then sorts through some of the rules on the merger. Besides, this paper introduces weighting parameters to adjust the location of cluster centers, thus closer to the centers of the actual. The introduction of weighting parameters can also processes outlier data point. Finally, the example shows the FCM algorithm based on entropy weighting is highly efficient.Secondly, the algorithm is applied to the analysis of the data telecommunications arrears. This section extracts the telecom on billing database, makes the data clean, and transforms algorithm data into a standard data format. Then there is the practical application of the algorithm. Through the form, we can find the characteristics of telecom delinquent customers, hich can provide reliable owe strategies.This innovation place and achievement lies in:(1) Propose a new fuzzy clusering algorithm based on information entropy, which reduces the dependence of algorithm on initial clustering center.(2) Introduce weighted parameters into the FCM algorithm based on information entropy, which further improves the initial clustering center position. (3) Apply clustering algorithm to the analysis of the data telecommunications arrears. On the one hand prove the practicability of the algorithm, on the other hand provide a theoretical for elecom operators.