SONG Chang-li1, JIN Lian-jie2, GUAN Feng3, LIN Mu-xi1*
1.School of Economics, Liaoning University, Shenyang 110036, China; 2.Transport Planning and Research Institute, Ministry of Transport, Beijing 100028, China; 3.School of Transportation Engineering, Shenyang Jianzhu University, Shenyang 110168, China
Abstract: This paper aims to establish an appropriate data model of the typical port cargo, and to make accurate throughput prediction. A multi-step hybrid prediction model is built by combination of the advantages of direct and iterative prediction mode based on support vector machine (SVM), and was used to empirically analyze 66 monthly cargo throughput data of Dalian Port. It was found that the average relative error between predicted by the multi-step hybrid prediction model and actual values was 5.7% for coal, 4.2% for oil, and 2.8% for container,with correlation coefficients of 95.5% for coal, 95.2% for oil, and 98.2% for container, indicating that the multi-step hybrid prediction model has accurate parameter calibration, and a relatively high level of prediction veracity. The findings indicate that the model is feasible to predict multiple cargo throughputs, which provides a strong technical support for port operators to grasp the overall port operation condition and to make operating decision.