Research Of Telecom Customer Quantity Forecasting Based On Data Mining

Nowadays, with the deepening reform of domestic telecom,development of communication technology and perfect of infrastructure, a new telecom market competition pattern has been formed. The competition pressure from both home and abroad makes operator of China Telecom realized that customer is the foundation of the development of enterprise. And the ways to attract, keep customer and make maximized interest became the focus which the operators concerned. The operation model of telecommunication operator transformed form drove by technique to drove by market and customer. And CRM is the sharp weapon that the telecommunication operators can use in the competition.Because of the increase of the choice offered the customer, the contention of the customer, especially the VIP customer become more and more fierce. It makes telecommunication operator face the unprecedented competition pressure. And the contention of the customer has became the focal point of the competition. Customer quantity are always the problems that telecommunication operator cares about most. Therefore the forecast of customer quantity becomes essential. It is the design basis for telecommunication operator to chart network development program and project design. It also can supply decision-making support to operator for formulating business development plan, improving management, making management decision.Author discusses the forecast of customer quantity in the background of some telecom operator's CRM in ChongQing. Now there are various algorithms, several commonly used algorithms are discussed in this paper. Using the sole algorithm has to brave some extent risk for the information the forecast need is from the history data, and the information that any single forecast algorithm attains is limited. For this reason, author obtain the forecast result from several algorithms by using weighted average of theirs. As the value ofσ2 can present the quality of one forecast algorithm, we can give each algorithm a weight according to it'sσ2.Finally, through using database of SQL Server2000 and Vc++6.0 programming realization proved the validity and feasibility of the algorithm. It can be used in practical application.