港口吞吐量预测的混合算法实证研究  

Empirical Study of an Intermix Algorithm for Forecasting Port Throughput

作　　者：蒋学炼[1,2] 吴永强[3] 史艳娇[2] 

机构地区：[1]天津大学建筑工程学院,天津300072 [2]天津城市建设学院天津市软土特性与工程环境重点实验室,天津300384 [3]中交第一航务工程勘察设计院有限公司,天津300222

出　　处：《中国港湾建设》2009年第2期7-11,共5页China Harbour Engineering

基　　金：国家自然科学基金项目(50779045)

摘　　要：提出一种新的港口吞吐量预测混合算法,基于天津港的历史数据进行了实证研究。首先采用主成分分析和相关分析从影响港口吞吐量的社会经济指标中甄别出显著影响因子,其次采用基于时间序列分析的三次指数平滑法获得其预估值,最后采用基于因果分析的遗传算法优化的神经网络法预测港口吞吐量。结果表明,影响因子预估中基准年的选择对预测结果影响很大,中长期预测应采用多个连续基准年的平均预估值,以平滑影响因子增长率的波动。A new intermix algorithm for prediction of port throughput is proposed and an empirical study was carried out on the basis of the historical information and data of Tianjin Port. Firstly, through using principal component analysis and correlation analysis, the most important factors are extracted from socioeconomic indexes which influence throughput of port. Secondly, the discreet values of these factors are obtained by employing the tri-exponential smoothing method based on time series analysis. Finally, the development of Tianjin Port throughout is forecasted using the artificial neural network method optimized by genetic algorithm technique based on causal analysis. The results indicate that the selection of base year has a great effect on predicting outcomes in estimating the values of factors. The discreet values of factors in medium-term or long-term forecasting should take the average values based on several sequential base years to smooth the fluctuation of the growth rate of impact factors.

关 键 词：吞吐量 主成分分析 三次指数平滑 遗传算法 神经网络 

分 类 号：U652.14[交通运输工程—港口、海岸及近海工程]