Predicting chaotic time series using relevance vector machine regression

A chaotic time series prediction method based on relevance vector machine regression (RVMR) is proposed in this paper. The RVMR model has a simpler model structure, a fewer number of control parameters, and a faster prediction speed with comparable approximate prediction accuracy in comparison with support vector machine regression (SVMR). In addition, the kernel function must not necessarily fulfill Mercer's conditions in RVMR model. Two typical chaotic time series, namely, Logistic and Hénon map are used to evaluate the RVMR's performance. The results show that the proposed method is effective in chaotic time series prediction.