An adaptive ensemble model of extreme learning machine for time series prediction

Time series predicting has become an important issue in many fields. The prediction methods which are based on the extreme learning machines have attracted many researchers. However, the predicted results of the extreme learning machines have some randomness. To obtain the better predicting performance and improve the randomness, we propose a new adaptive ensemble model of extreme learning machines (Ada-ELM) in this paper, which can adjust the ensemble weights automatically. We test Ada-ELM on two actual time series comparing with other four prediction methods based on the extreme learning machines. In the experiments, Ada-ELM outperform the other four methods and shows the good adaptability of our prediction method to different time series.