Risk support vector machine for predicting the trend of enterprises development

Enterprises development trend depends on many factors such as product, distribution, manpower and market. These factors are interactive and coupling in statistic data that makes it difficult to determine which enterprise is promising and which one should transfer type. How to classify the enterprises condition and predicate their future is urgent to solution. This paper utilizes the inner production kernel function to extract the useful nonlinear information and eliminate the redundant data from statistic information. Then present a risk support vector machine to improve the classification capability of multiple variables system under disequilibrium and limit samples. Through introducing the risk probability, we can focus on the important feature and classify the enterprise with high precision, then invest the promising enterprises to realize the high-tech innovation in market. Application of Beihang Discovery Park indicates that the risk support vector machine not only can solve the problem of nonlinear feature extraction but also can realize the optimal predication classification under unbalanced and small samples with high classification precision.