Relevance Vector Regression Multi-step Prediction Method Optimized By Particle Filter

In order to weaken cumulative error in multi-step regression prediction effectively, relevance vector regression multi-step prediction method optimized by particle filter is proposed in this paper. Relevance vector machine is used to build regression prediction model of characteristic parameters of state. Regression model is acted as system state function and probability distribution of prediction output is acted as importance distribution. By using particle filter algorithm, prediction data is corrected dynamically and the optimal state is estimated. It is proved that the precision of relevance vector regression multi-step prediction method optimized by particle filter is improved by contrast experiment.