Research On Clustering Algorithm And Application In Telecom Revenue Assurance System

The rapid development of information technology and economic globalization makes the current telecom operators facing a complex and changing external environment, and the resulting series of challenges and opportunities of making the pressure in the telecom operators rapidly growing. In order to increase business profits and improve operational efficiency, carriers need a complete specification of the Revenue Assurance system to support business processes and business systems to monitor, look for the loss of point, to stop or prevent revenue loopholes. Data Mining is to comply with the data from the massive extraction of useful knowledge in this demand came into being developed data processing techniques, is a key step in knowledge discovery, a lot of the technology makes it based on statistics found at a deeper level within the data link have unique advantages, the development of hardware computing power for data mining provides further support.This work mainly includes the following:(1)Overview of clustering technology and telecommunications system of Revenue Assurance and historical context, the present studies is described, including the basic concepts of clustering and the basic theory of Revenue Assurance. On this basis, focusing on the traditional clustering algorithm to do a more detailed description, and pointed out the existing problems, but also the analysis of these issues. Also introduced the Revenue Assurance system, the telecommunications architecture and clustering technology in the protection of telecommunications Revenue Assurance.(2)Partition-based clustering technology for the number of clusters and the initial class in the choice of heart problems, this paper, a slow start based on fast adaptive clustering algorithm. The algorithm uses Voronoi centroid to improve the choice of the initial centroid, and include the use of a class of density, the distance between classes, a class composed of multiple factors including tightness of the decision function to make the assessment of the clustering effect can be intuitive and more accurately reflect the effect of clustering. Algorithm from a large number of clusters begin to function value judgments, followed by exponential decrease of the number of clusters were observed function value judgments to reach inflection point when the decision function, then a linear decrease of the number of clusters, when the decision function value again reach the inflection point, the number of clusters at this time determine the best number.(3)Designed a single audit CDR system for Revenue Assurance, which uses MFC language development, intuitive interface with a graphical display of the data mining results. While the system in a timely manner convenient to modify all the parameters, comprehensive analysis of the experimental data. Then use the telecommunications system 600000 if a single record of the test, by observing the decision function value found suitable for the data if a single decision factor. The results show that the system has good operability. This article summarizes the research results the last, points out the deficiencies and proposed improvements.