An effective high-quality prediction intervals construction method based on parallel bootstrapped RVM for complex chemical processes

Abstract

Data-driven techniques have been becoming increasingly popular and widely used for prediction in complex chemical processes. In general, prediction results are usually provided with point estimations. However, point estimations cannot meet the requirement of accuracy due to the characteristics of high-dimension, high nonlinearity, and containing noise of process data. In order to deal with the trend and the uncertainty of process data, an effective prediction intervals (PIs) method based on bootstrap and relevance vector machine (Bootstrapped RVM) is proposed in this paper. In the proposed method, bootstrap is adopted to obtain PIs and RVM is used as a regression tool. In order to accelerate the training and testing phases, a parallel algorithm is utilized in the proposed Bootstrapped RVM method. In addition, to better evaluating the quality of PIs, some performance indicators are improved. Finally, the proposed method is validated by using a standard function and High Density Polyethylene (HDPE) data. Compared with some other PIs methods, the simulation results show that the proposed method can achieve better performance in terms of prediction accuracy and training time.