VMD-HHO-DELM: Decomposition and Optimization Model with Based Deep Extreme Learning Machine Algorithm for Predicting Short-Term Traffic Flow

Abstract

The traffic flow prediction solutions are crucial method to optimize transportation efficiency in an urban management and intelligent transportation system (ITS). However, accurate prediction of non-stationary traffic flow data is a challenging task. This paper proposes a novel deep extreme learning machine (DELM) model coupled with the variational mode decomposition (VMD) and Harris Hawks optimization (HHO) algorithm to predict short-term traffic flow(STTF). Firstly, considering the effects of unsteady traffic flow data, we design the VMD algorithm to decompose the intrinsic mode functions (IMFs) obtained from the unsteady data signal. Secondly, we design DELM, which combines the learning effectiveness of extreme learning machine (ELM) with the deep neural network framework of stacked autoencoder (AE). In additional, to improve the performance of our model (named VMD-HHO-DELM), we design the HHO algorithm to optimize it for a more efficient and accurate model. Extensive experiments on steady and non-steady data sets demonstrate that our proposed method has significant effects compared to others. The R2 score is as high as 0.9984 and 0.9993 for prediction of non-stationary and stationary data, respectively.