基于$\varepsilon $-支持向量机回归的快速公交到站时间预测

Abstract

选择$\varepsilon $-支持向量机回归($\varepsilon; $-SVR)算法预测快速公交(BRT)车辆的到站时间,以提高公共交通的准点性.分别对BRT的停靠时间和路段行驶时间建立模型.根据分析,在停靠站时; 间预测建模过程中选取车头时距、时段、天气等7维特征向量作为模型输入,采用人工调查法,对厦门BRT-1路的数据进行采集,归一化处理后建模.仿真结果; 显示该模型能够比较准确地预测厦门BRT-1路的运行路线到站时间,并验证天气因素对该线路的到站时间预测影响最大.

In this article,we select $\varepsilon $-support vector machine; regression ($\varepsilon $-SVR) algorithm to predict the bus rapid; transit (BRT) arrival time,in order to improve the public transport on; time.The bus stop time and road travel time models are; established.During the modeling process,seven dimensional features such; as the headway, time, weather etc.,are chosen as model inputs.; Artificial investigation method is used to collect the data of Xiamen; BRT-1. These data are normalized during the modeling process. Simulation; results show that the model can accurately predict bus running time of; Xiamen BRT-1. Results verify that weather factors exert the highest; influence on the arrival time of Xiamen BRT-1.