The Study And Application On The Prediction System Of The Behavior To Telecommunication Customer Cheats

Today, reforming with correspond by the communication technical development and foundation facilities developments perfect along with the reorganization of the telecommunication industry, each telecommunication company launched the vigorous competition with business etc realm at the customer. It is traditional to regard producing as the center and regard selling the product as the market strategy of the purpose to is regarded customer as gradually center, regard serving as the market a focus for replacing, customer's resources becoming business enterprise competing of the purpose. In customer's resources competition war, each telecommunication company's market competition caused into net threshold lowered ,it gave the customer the real benefits to draw on customer. In all telecommunications customer, although the malice owes customer's community occupied the total community only very small a part, the loss of the this group that customer resulted in the telecommunication industry is enormous, each telecommunication company in order to avoid, save or alleviate this part of losses have to adopt every kind of measure. From the point of this article, we established a model of the prediction of the behavior to customer cheats according to cheat-customer's the history communication behavior in customer analysis. In fixed telephone correspondence business, now the strategy that owe an adoption is to owes behind the first month of fee practice to limit the call sign, namely the customer's can be shouted but can't be active to exhale; Owing behind the second month of fee practices the more scathing restrict, that customer can't exhale nor to is shouted; The third month practice to dismantle the machine compulsory. Obviously in here can't owe above fee appearance 1 month or 2 customers of month definitions are for cheating the customer, but a customer for can thinking that quilt compulsory dismantling machine just cheats the customer really. Making use of the RBF nerve network sets up the mold, passing the single history in converse data in customer data train, establishing the customer belong to the serve to belong to the sex to consume the estimate model of the data with the customer. Belong to the serve the customer community that belong to according to the customer that predict dangerous degree in different from the customer consumes the data, can compute a dangerous degree for, making use of the cluster technique according to customer that customer cheat divide the line the customer grade in group. Customer group that aim at the different and dangerous degree adopts the different strategy, aiming at to cheat for example the dangerous and high customer group can adopt to urge to pay the strategy scathingly, limitting for example the service, biggest degree shorten to pay a buffer time etc. Finally,The paper makes use of the VC++6.0 proceeded the design of the man-machine interface and the system procedure.