Abstract: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 algorithm’s outcome is
another algorithm’s 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.