基于最小二乘向量机的公交站点短时客流预测  

Prediction of Short-term Passenger Flow on a Bus Stop Based on LSSVM

作　　者：郭士永[1] 李文权[1] 白薇[1] 张东[1] 

机构地区：[1]东南大学交通学院,南京210096

出　　处：《武汉理工大学学报（交通科学与工程版）》2013年第3期603-607,共5页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家重点基础研究发展(973计划)项目(批准号:2012CB725402);国家自然科学基金项目(批准号:50978057)资助

摘　　要：考虑上下游公交站点、历史同期客流和相邻间隔输入因子β三者的影响,采用最小二乘支持向量机回归算法建立预测模型,并利用粒子群算法优化模型参数.实例验证结果表明:三者均会对预测精度产生影响;当β=3并在多输入变量中设有上下游站点、历史同期客流维度时,该预测模型相比预测性能最好,平均绝对误差为0.625 0,均方误差为0.914 5.The paper considered the impact of bus stops in the upstream and downstream,the historic passenger flow in the same period and adjacent interval input factor β,adopted Least Squares Support Vector Machine to establish the prediction model,and utilized particle swarm algorithm to optimize LSSVM parameters.Finally,the proposed forecasting method was validated in a instance.Prediction results suggested that the mentioned three factors have impact on forecasting accuracy,and that when β=3 and when bus stops in the upstream and downstream,the historic passenger flow are both taken into account,the performance of the prediction model with mean absolute error=0.6250 and mean square error=0.914 5 is the best of three models.

关 键 词：城市公共交通 短时客流预测 最小二乘支持向量机 粒子群优化算法 

分 类 号：U491.14[交通运输工程—交通运输规划与管理]