Research And Application Of Call-back Service Analysis Based On Data Mining

In recent years, the number of mobile users increased rapidly and the number of landline users is relatively shrinking. Because the AUPA value of mobile users is much higher than the value of landline users, the operating revenue gap between China Mobile and other operators is gradually increased. Due to the unequal competition between mobile networks and fixed networks, some landline operators and virtual operators make use of call-back service to snatch the long distance charge of mobile network.There are two traditional ways to find call-back service. One way is to do online search for advertisement of call back service on website and find the callback service number. Another way is to investigate the telecommunications business card trading market. However, the traditional ways consume a lot of resources and the result is unsatisfactorily. So, how to make use of the data mining technology to analyze and forbid the call-back service numbers from huge network information is imminent for mobile operators.In this thesis, in order to achieve efficient analysis and processing of call-back service, we propose to construct a signaling acquisition system to collect the related signaling message of call-back service, and make use of the data mining technology to analyze the data and achieve efficient result. The work we have done is as following:(1) First, the network structure and interface between network elements of Hunan Mobile, signaling protocols of GSM network and technology basis of call-back service is introduced.(2) Via in-depth study of call-back service and analysis of related signaling processes, signaling monitoring platform is constructed to collect the call-back service related signaling, network interface data in GSM network. The collected information is used as the original data source of call-back service analyzer.(3) Via the analysis of association rule mining algorithm, lots of original source data are collected for analyzing, modeling and processing. The analysis result can provide timely and accurate information to network maintenance personnel to make decision.(4) Based on the result of data analysis, linear regression covering algorithm is used for coverage testing, then through the classification of operators and time intervals, we find characteristics of call-back service numbers, we get the maximum accuracy. At the same time, the rules of statistical analysis, and Bayesian probability analysis are used to improve the analysis tools. The analyzer after optimization can provide nearly 10% increase of probability of correct handling.