Selection Method and Application of Kernel Parameters
Xunfang LIU
Hunan City University, Yiyang, Hunan, 413000, China
Abstract: The use of relevance vector machine (RVM) Gaussian radial basis kernel function regression model, the relationship between the nuclear parameters and model performance is complex. Aiming at the problem of how to determine the kernel parameters of RVM, put forward a kind of based on the AIC criterion to select kernel parameter of RVM. First, based on Akaike Information Criterion (AIC) thought, come up with a new statistic Q, while the Q as a fitness function. And then use the differential evolution algorithm (Differential
Evolution Algorithm, DE) on the kernel parameter optimization, in order to choose determine the kernel parameters. Finally use the algorithm to establish RVM regression model to short-term predict gold price. Experimental results show that the prediction model than the traditional method to establish the model has higher precision and better fitting the generalization ability, and further demonstrate the feasibility and effectiveness of the AIC-based criteria for selecting RVM kernel parameter method.
Keywords: Radial Basis Function, Kernel Parameters, Relevance Vector Machine, Differential Evolution Algorithm, AIC Criterion, Gold Prices