基于特征选择和神经网络的铁路货运量预测  

Prediction of Railway Freight Volume Based on Feature Selection and Neural Network

作　　者：段力[1] 王开鹏 刘聪健 王孙超 Duan Li;Wang Kaipeng;Liu Congjian;Wang Sunehao(School of Civil Engineering ＆ Mechanics,Huazhong University of Science ＆ Technology,Wuhan 430074,China)

机构地区：[1]华中科技大学土木工程与力学学院,湖北武汉430074

出　　处：《物流技术》2018年第9期52-57,共6页Logistics Technology

基　　金：国家自然科学基金(51208222);国家社会科学基金重大项目(13&ZD175)

摘　　要：首先论述了传统铁路货运量预测方法的不足,然后介绍了广义回归神经网络预测模型以及将数据输入基于广义回归神经网络预测模型之前,采用数据筛选和模糊聚类相结合的特征选择方法,最后以广州为例,对该方法进行了验证。In this paper, we first recounted the shortcomings of the traditional forecasting methods of railway freight volume and then introduced the general regression neural network forecasting model as well as the feature selection method which combined data screening and fuzzy clustering before inputting the data into the model. At the end, the validity of the method was demonstrated in the case of Guangzhou.

关 键 词：铁路货运量 预测 特征选择 神经网络 

分 类 号：F224.0[经济管理—国民经济] U294.1[交通运输工程—交通运输规划与管理]