Container throughput forecasting using a novel hybrid learning method with error correction strategy

Port throughput forecasting is not only a complicated problem but also a challenging task in port management fields. In this study, a novel hybrid learning model that utilizes effective decomposition techniques, such as variational mode decomposition (VMD), machine learning, optimization algorithms, and error correction strategies, is developed for container throughput forecasting. First, VMD is adopted to divide the original data into a finite set of components: subsequently, the different features hidden in the container throughput time series can be extracted by different modes from low frequency to high frequency. Next, each component obtained from VMD is modeled and predicted by an extreme learning machine (ELM) technique optimized by a butterfly optimization algorithm (BELM): subsequently, another BELM predictive model based on a training error series is constructed to predict the consequent error. Next, the correction of preliminary prediction values is calibrated. Finally, hypothesis testing, six model evaluation criteria, eleven comparison models, and two case studies are utilized to comprehensively evaluate the developed hybrid method. Based on the experimental results and related analysis, it can be fund that the proposed hybrid method is superior to the eleven comparison models in terms of prediction accuracy, and can be regarded as an effective tool for port operation management. (C) 2019 Elsevier B.V. All rights reserved.