Reasearch On The Application Of Combination Forecast Model In The B2B’s Risk Client

With the rapid speed development of China’s Internet and promoting the informationization process in the industries, electronic Business continues to keep a rapid growth this year.In2011the revenue for B2B e-commerce of our country reached13billions yuan.The one major reason for the growth of revenue for B2B E-Marketplace in2011is medium-sized and small enterprises increased investment in e-commerce and increased use of value-added services.The other reason is core operators adopt effective measures to improve the quality of service and expand the scope of business.While increasing investment in e-commerce and improving the quality of service will find that China’s e-commerce platform will appear speculation stage and low ebb, some problems such as electronic business survival, the technical implementation and so on cover the credit and the credit risks, and the low frequency of online trading also makes the risk less obvious.This paper is based on the three forecasting model methods and the optimal combination forecast model for B2B e-commerce platform risk customers.The ultimate goal of this paper is to establish an optimal and the most effective risk customer warning model system for B2B E-business platform.Based on the three single prediction model such as the BP neural network, decision tree learning method and Logistic regression model,the prediction model of index is set up.And then through the Kendall union coefficient consistency inspection of three kinds of methods to do the optimal combination forecast model under the Kuhn tucker constraint conditions. Finally it gets that the optimal combination forecasting model is able to make the effect better than other single prediction models and then puts forward some solutions to warning customer management under risk state.This paper has two innovations. Firstly, it is the first time to use combination forecasting model to forecast risk customers in B2B E-commerce platform and good effect is get. Secondly, Kendall Concord coefficient consistency test is taken before combination forecasting. It is a less scientific method for the past scholars to take direct forecast.Because the B2B industry which is the media to build linkages and transactions between enterprises and lack of physical behavior record of transactions, it can not monitor the trading behavior of customers or obtain certain information from the perspective of the trading behavior of the risk analysis,this brings great challenge to the risk analysis and identify.Another advantage of the optimal combination forecasting model mentioned in this paper is that the model can be improved as time went on.Nowadays the research of forecasting the risk customers of B2B E-commerce platform is still single forecast.In the future,we can try to combine the four forecast methods and correct new index to accommodate new things.