The Designation And Implementation Of Key Mobile Telephone Client Leaving-Network Alarming Model

For telecom operators, to attract and keep customers is the foundation to ensure telecom operators to survive and develop, while to fully dig the potential profitability of clients is the key to enhance the core competitiveness. There is the principle of 20/80 in telecommunications market. In other words,80% of the profits of enterprise come from 20% of the large clients. Large clients become the focus to be shot for by mobile communications enterprises. But, the cost to develop new large clients is far greater than that of retaining current large clients. Therefore, to reduce customer loss is the strategic task for mobile operators and if the possibility of leaving-network large clients can be alarmed in time is the key to complete the strategic task.To research and implement the leaving-network alarming model is the center in this dissertation. Firstly, some typical data mining algorithms to predict are introduced, such as Bayes algorithm, neural networks, clustering algorithm. Secondly, combining with customer relationship management theory and data mining standard process, the leaving-network alarming model for key clients is proposed based on Bayesian network. According to the key clients'action characteristics in the network, the factors of lients' leaving-network are analyzed and the probability of clients' leaving-network is calculated using Bayes data mining algorithm. Then, the key clients leaving-network alarming system is designed and realized based on the model. The network architecture, software architecture, databases, and the core algorithm of the system are described in detail. The system performance is analyzed through the applications and the evaluation. Analysis showed that the proposed model can successfully alarm the possible leaving-network large clients and play a key role to maintain the mobile operator's core competitiveness. At the same time, the systm can effectively reduce the operating cost of company. The model's validity and the system's utility are also proved.