摘要
提出一种适用于多影响因素回归拟合的全社会客运量预测的点积-平移型支持向量机算法.该算法能够全面、系统地分析影响客运量需求变化的关联因素,通过关联元素去冗处理,确定作为支持向量机算法输入变量的核心关联元素.考虑到客运量预测是一个基于时序变化的外推过程,并且受社会经济发展中多项影响因素的制约,数据变化存在着阶段性特征,提出利用点积-平移型核函数来拟合需求变化过程.对历史数据集的测试结果表明,该算法性能评价满足要求,可为远景客运量预测提供理论依据.
A dot product-translation support vectors machine algorithm was proposed,which applied to whole social passenger traffic forecast with multiple influencing factors regression.This algorithm can analyze the associated factors of influencing passenger traffic demand variation roundly and systematically.After reducing redundancy,the key associated factors are determined as input variables of support vectors.Considering passenger traffic forecast as an time series-based extrapolated process,which is restricted by multiple factors in social and economic development,and data can be changed with stage characteristics,the dot product-translation kernel function proposed is used to match the demand variation process.Test result of history data indicates that the performance evaluation can meet the requirement,which can provide a basis for passenger traffic forecast with distant view.
出处
《大连海事大学学报》
CAS
CSCD
北大核心
2012年第4期99-102,共4页
Journal of Dalian Maritime University
基金
"十一五"国家科技支撑计划(2006BAJ18B01-03)
关键词
支持向量机
去冗处理
点积-平移核函数
客运量预测
support vectors machine
reduced redundancy processing
dot product-translation kernel function
passenger traffic forecast