Investigation of diversity strategies in RVFL network ensemble learning for crude oil price forecasting

Abstract

To address the drawback of single machine learning prediction model which cannot capture the complex hidden factors of crude oil price, ensemble learning method has been widely verified as an excellent solution for crude oil price forecasting. In ensemble learning, diversity strategy is one of the most important determinants to obtain good performance. So, this study introduced the promisingly efficient and fast RVFL network as base models to the framework of ensemble learning and explored diversity strategies in the proposed RVFL network ensemble forecasting model to obtain good performance. Specifically, the impacts of five different strategies including data quantity diversity, sampling interval diversity, parameter diversity, ensemble number diversity and ensemble method diversity on the performance of RVFL network ensemble learning have been examined and analyzed. Experimental results found that the accuracy of ensemble learning models would be increased if diversity strategies were carefully selected. Moreover, the proposed multistage nonlinear RVFL network ensemble forecasting model was consistently better than that of single RVFL network model in terms of the same measurements.