Abstract: Accurate and real-time travel time information of buses can help passengers better plan their trips and minimize waiting time. A dynamic prediction model for bus travel time is proposed in this paper, based on Support Vector Machine（SVM） and Kalman filtering-based algorithm. In the proposed model, the well-trained SVM model predicts the baseline bus travel time from the historical bus trips data; the Kalman filtering-based dynamic algorithm can adjust bus travel time by using the latest bus travel information together with estimated baseline travel time. The dynamic model is tested with the data of number 223 bus route in Shenzhen city. The results of prediction accuracy among the proposed dynamic model, pure SVM model and Artificial Neural Network（ANN） model are compared based on the empirical data. Results show that the proposed dynamic prediction model for bus travel time based on Support Vector Machine and Kalman filtering-based algorithm has better prediction accuracy and dynamic performance.

Key words: Support Vector Machine（SVM）, Kalman filtering, artificial neural network, bus travel time, dynamic prediction

摘要： 准确以及实时的公交车行程时间信息能够帮助出行者更好地规划行程，减少出行者的等待时间。提出了一种基于SVM-Kalman滤波的公交车行程时间动态预测模型。模型中，经过良好训练的SVM模型从历史数据进行预测得到行程时间基准；Kalman滤波动态算法在基于SVM模型预测值和最新公交出行信息的基础上对结果进行调整。以深圳市223路常规公交线路为实例，将动态模型的预测精度结果与单一SVM模型、ANN模型结果进行对比，结果表明，基于SVM-Kalman滤波的公交车行程时间动态预测模型的预测精度更高、动态性能更好。

关键词: 支持向量机, 卡尔曼滤波, 人工神经网络, 公交车行程时间, 动态预测