Research On Frequent-pattern Mining Technology And Its Application On Revenue Assurance Systems

With the increases of openness and competition of the domestic telecommunication industries as well as the development of new developing computer and communications technology, telecommunications market is rapidly expanding and 3G era has also brought with its new opportunities and changes to the telecom market. Therefore, we apply data mining technology to the retrieval of income loopholes in business operations and patch them, which lead to reducing the loss of operating revenue, improving profits level, standardizing business to guarantee a sound revenue assurance system. Thus, it becomes a major goal operators are pursuing. Based on revenue assurance system, this paper researches on mining frequent-pattern of data mining techniques, which mainly include:Firstly, research on frequent pattern mining technology and telecommunications revenue assurance system is overviewed, including the basic concepts of frequent pattern mining techniques and the basic theory of revenue assurance. On this basis, further information on the classic frequent pattern mining algorithm and income security systems of telecommunications architecture is introduced. Frequent pattern technology in application of telecommunications income security is highlighted.Secondly, In order to solve the problem that minimum support threshold can not be predetermined in frequent pattern mining process, this paper presents a new frequent pattern mining algorithm to get the optimum support threshold. Algorithm by linear combination of support threshold, the number of frequent items and frequent number, generates a decision function which can evaluate the effect of frequent pattern mining. It is more intuitive and accurate to reflect the frequent pattern mining results.This decision function selects a smaller support threshold as an initial threshold at first, and then exponentially increases value of support.Decision function results are calculated under different values of support. Values of support are increased linearly when the results of decision function achieved inflection point. When the decision function arrives at the inflection point again, then the threshold before arriving at the inflection point is judged as the optimum support threshold. Experimental results show that the new algorithm performs better than the Apriori algorithm for mining frequent patterns mining results.Thirdly, a small CDR phone bill audit system of telecommunication revenue assurance is designed and completed.600,000 telecommunications CDR phone bill data is mined. By adjusting the parameters of the decision factor of the function, the experimental data is analyzed comprehensively, and the actual mining effect is assessed and compared comprehensively. The results show that the system has good operability.