A Data Mining Approach to Modeling the Behaviors of Telecom Clients

We begin this paper by describing our rationale and overall design of an exceptional client model based on data mining algorithm. Then, we continue by summarizing the simulation details and describing the type of results obtained from implementing the proposed system, which consists of three heterogeneous data mining algorithms. The idea behind the model is that three heterogeneous data mining algorithms are combined in the phase of data processing and one algorithmpsilas outcome is another algorithmpsilas input. By using the method of neural network, the exceptional client attribute weight is calculated based on the original dada. The characteristics of exceptional client are identified accordingly by using the method of decision tree based on the attribute weight. Then, the distinguishing model is generated adaptively on the basis of clustering. The combination of the three algorithms helps distinguish exceptional client effectively. It can not only alarm the existing system but also analyze and specify the data of exceptional client and consequently supports the distinguishing system.