Online Sequential Extreme Multilayer Perception with Time Series Learning Machine Based Output Self Feedback for Prediction

This study presents a time series prediction model with output self feedback which is implemented based on online sequential extreme learning machine. The output variables derived from multilayer perception can feedback to the network input layer to create a temporal relation between the current node inputs and the lagged node outputs while overcoming the limitation of memory which is a vital port for any time-series prediction application. The model can overcome the static prediction problem with most time series prediction models and can effectively cope with the dynamic properties of time series data. A linear and a nonlinear forecasting algorithms based on online extreme learning machine are proposed to implement the output feedback forecasting model. They are both recursive estimator and have two distinct phases： Predict and Update. The proposed model was tested against different kinds of time series data and the results indicate that the model outperforms the original static model without feedback.

Adaptive forgetting factor OS-ELM and bootstrap for time series prediction

Online sequential extreme learning machine(OS-ELM)for single-hidden layer feedforward networks(SLFNs)is an effective machine learning algorithm.But OS-ELM has some underlying weaknesses of neglecting time series timeliness and being incapable to provide quantitative uncertainty for prediction.To overcome these shortcomings,a time series prediction method is presented based on the combination of OS-ELM with adaptive forgetting factor(AFF-OS-ELM)and bootstrap(B-AFF-OS-ELM).Firstly,adaptive forgetting factor is added into OS-ELM for adjusting the effective window length of training data during OS-ELM sequential learning phase.Secondly,the current bootstrap is developed to fit time series prediction online.Then associated with improved bootstrap,the proposed method can compute prediction interval as uncertainty information,meanwhile the improved bootstrap enhances prediction accuracy and stability of AFF-OS-ELM.Performances of B-AFF-OS-ELM are benchmarked with other traditional and improved OS-ELM on simulation and practical time series data.Results indicate the significant performances achieved by B-AFF-OS-ELM.