Application Of Data Mining In Mobile Communication

As China's domestic telecommunications reform deepens, the WTO accession, mobile communication Business of competition is becoming increasingly fierce. Compared with other industries, mobile communications Operators have more relevant user data. Who can rightly excavation and analysis of these data implied in the knowledge, who would be better able to provide products and services users can find more opportunities to win in the competition. Mobile communications is the leading domestic enterprise data warehouse system for the implementation of the user based on data mining users integrity analysis to provide the conditions. The projected users integrity analysis system operating system as an important component of establishing customer loyalty through the prediction model, and the enterprises will be able to detect the network from the user, to take timely measures to reduce the occurrence of users from the network. Therefore, the integrity of the Prediction of users of mobile enterprises to reduce operating costs and improve operational performance has a very important significance.China Mobile in 2001 completed the decentralization BOSS (Business Operation Support System, that is, operations support systems) building, and a large number of users BOSS data storage, user data and user communications consumer behavior information and data. China Mobile from the beginning of the 2001 BOSS nationwide based on the data warehouse construction and preparatory work and the preparation of a specific guidance norms, a national operating the unified system of the building work. In 2002 the provincial operation analysis system have started in 2003 has been completed nationwide data warehouse building. China Mobile operating system based on business analysis and management needs, data warehouse, data mart to establish suitable for enterprises to provide a unified view of data for the conduct of the user based on data mining loyalty analysis and the establishment of customer loyalty forecast model provides data mining basis. For mobile enterprise development needs and the characteristics of these data for the reorganization of the structure by more conducive to the point of decision analysis to rearrange and organizations, will enable mobile enterprise data into a truly valuable information, as well as enterprise development and provide a basis for management decision-making.According to data mining technology to business objectives and the established existing problems of a large number of operational data mining, which reveals the hidden laws, and its modeling, guiding and applied in actual business. Faced with a deluge of data, using the correct operation of the telecommunications market thinking, using data mining techniques to explore hidden in operational data users and the high value-added information resources and achieve customer loyalty analysis and forecasting will be a telecommunications enterprises in the information age invincible in the competition in the market and the key. The reserve data within enterprises and enterprises outside the fierce market competition led to the data mining technology based on the analysis of customer loyalty research inevitable. Mining data value, and promote knowledge management and intelligent decision-making telecommunications business enterprises enhance overall competitiveness of the inevitable choice.Based on the mobile communications business status quo, a comprehensive analysis of the data mining technologies and data mining enterprises in the application of communications, the proposed based on competition, evolutionary computation of customer loyalty prediction algorithm, and the establishment of customer loyalty forecast model. Major research include:1. summarizes the data mining technology, including data mining process, data mining principal methods, and data mining technology in mobile communication applications, data mining proposed in the settlement of the mobile communications business forecasts, user profiling, high-end user recognition, customer loyalty analysis , the breakdown of user groups such issues as the main method and the use of technical means.2.against competitors marketing policies customer loyalty issues raised in King Under the premise of an strategy of customer loyalty prediction model and experimental analysis were compared.3.the analysis of the evolutionary computation suitable for solving optimization problems on the basis of evolutionary computation made by the customer loyalty prediction algorithm developed by evolutionary computation of customer loyalty analysis model and experimental analysis were compared;Evolutionary Computation need to convert the data for discrete data. Customer loyalty involved in the forecast of some type of user information are numerical data. Could be separated, grouped into categories of data. For discrete data in the process of distribution of the original data remain under the premise of the numerical classification of data processing pretreatment process, the process for discrete data by the text of the application ECCA algorithm customer loyalty forecast of the whole process are very important. In introducing kmeans algorithm based on the application of self-organizing data distribution algorithm for solving discrete ways to deal with the k value kmeans algorithm set a direct impact on the issue of classification results. Distributed by the self-organizing algorithm neurons formed m, m is the maximum number can be divided into the number of interval types. Attribute values as input algorithm in the process of self-organization, the right neurons constantly being updated gradually to the real attributes of the initial classification close, until a growing number of neurons no longer updated, completed the continuous data discretization process. Mobile communications enterprise data sources in the number of attributes, some attribute the strong correlation between the presence of a discrete and the data sets, there are some of the problems did not affect the decision-making redundant attributes. In order to improve the efficiency of data analysis algorithms, the data sets must first choose to attribute to identify a decision-making capacity and the original data set the same minimum attribute set. The high-dimensional data against data analysis or data mining algorithm complexity of the time dimension with the growth of spending time with the issue rose by appropriate feature selection methods can reduce the dimension of data, and can maintain the ability to distinguish the original data. We used chi-square statistic for the correlation of quantitative attributes the results, according to the card table found independence confidence levelα. For a certain subset of the attributes, according to twoαis an orderly sequence, a sequence is classified all the attributes and properties of alpha orderly sequence, and the other sequence is the light of all the attributes and properties of alpha orderly sequence. Using various attributes of the two sequences in the potentiometer (position diff) choice attributes. The proposed based on potentiometer attribute selection algorithm FSBPD, the algorithm can be used to maintain data in the original distribution of cases, removed from the data for decision-making on unrelated redundant attributes. Finally, the theoretical analysis algorithm and experimental results are presented and analysis, laboratory tests show that the algorithm has a good attribute FSBPD choice capacity.4.the selection of data mining algorithms decision tree method of customer loyalty forecast .