Gray Compensating RBF Prediction Model Based on Structural Risk

A new prediction model that combining the merits of support vector machine (SVM) and gray RBF neutral network is proposed in this paper. First apply structural risk minimization principle to optimize the modeling method of RBF neutral network, so that the radial basis centers and network weights could be acquired directly. Then use error compensator of RBF neutral network based on structural risk to compensate the predicting results of GM (1,1) model. The comparative experimental results show that this model is capable of improving the data predicting accuracy, as well as the generalization ability of neutral network.

[1]  Bernhard Schölkopf,et al.  Comparing support vector machines with Gaussian kernels to radial basis function classifiers , 1997, IEEE Trans. Signal Process..

[2]  Tang Wan-mei.  New forecasting model based on grey support vector machine , 2006 .

[3]  Yang Yu,et al.  The Dynamic Grey Radial Basis Function Prediction Model and its Applications , 2006, First International Conference on Innovative Computing, Information and Control - Volume I (ICICIC'06).

[4]  Luo Zhong,et al.  A study on grey RBF prediction model , 2005, 2005 International Conference on Machine Learning and Cybernetics.

[5]  Deng Fang-ping,et al.  Introduction to Model selection of SVM Regression , 2006 .