基于多源数据融合的干线公交车辆行程时间预测

为提高城市中心区干线公交车辆行程时间的预测精度，在拟合公交车辆行程时间分布特征的基础上，提出基于多源数据的干线公交行程时间预测模型.对RFID及GPS检测器获取的实际数据进行预处理及分布拟合，其中混合高斯分布函数适用于单路段拟合，对数正态分布适用于多路段的拟合.采用皮尔逊相关性系数对影响行程时间的因素进行相关性分析，其中上游路段前2 个时间窗的平均行程时间的影响最大.分别采用ARIMA、改进的SVM模型对行程时间进行预测，其中改进的SVM模型的平均绝对百分比误差为6.26%，优于ARIMA模型的11.69%，更适用于短距离交叉口间的公交车辆行程时间预测.

Travel Time Prediction of Main Transit Line Based on Multi-source Data Fusion

In order to improve the travel time prediction accuracy of the main transit line in the city center, based on the analysis of the distribution characteristics of the bus travel time, the method of forecasting the transit time based on multi- source data is proposed. The actual data obtained by radio frequency identification and global positioning equipment are preprocessed, and the fitting analysis is carried out by mathematical statistics model. The mixed Gaussian distribution function is suitable for single-segment fitting, and the lognormal distribution is suitable for multi- segment fitting. The Pearson correlation coefficient is used to sort the time series and spatial factors affecting the travel time. The average travel time of the first two time windows of the upstream road segment has the greatest impact. The travel time is predicted by using ARIMA and SVM models, respectively. The mean absolute percentage error of the ARIMA model is 11.69%, and the mean absolute percentage error of the support vector machine model is 6.26%.