(The) development of the value-added service prospect scoring model in mobile communication market using data mining techniques

Abstract

When it comes to the situation of the mobile communication market, the growth rate of new customers’ subscription is getting mature, which is due to the fierce invitation competition so far. Moreover a lot of civil communities intend to lower the calling service price as well as the mobile data service one, which is based on the high profitability of the mobile communication service providers. At long last the authorities concerned decided to lower the mobile calling service price. The mobile communication service providers are inevitable to find the breakthrough, not of the additional new customer subscription, but of the new revenue sources, which leads to find the additional revenues based on the attraction of existing customers. One of the alternatives is the expansion of a charged value added service. The wide use of a charged value added service can bring both the new revenue source and the protection of churners. The objective of this study is to develop the value added service prospects scoring model, that is, to execute target marketing based on the customers’ preference to value added service. The target marketing can reduce a cost of marketing and improve the marketing efficiency and effectiveness. This study experimented the scoring model with the data of value added services including the logistic regression model, the artificial neural network model, and the decision tree model. Each models is applied to service A and B respectively. The criteria of model assessment are the hit ratio of validation set and lift value. When it comes to the hit ratio, the output of scoring model for service A reveals that the artificial neural network outperforms the others, so does the scoring model for service B. In the case of lift value, however, the logistic regression outperforms the others. Thus this study gets the second stage to consider the integrated scoring model to overcome the instability of a single model. This study decides the integrated scoring mode...