组合非参数回归和卡尔曼滤波的公交车到站时间预测

Bus Arrival Time Prediction Combining Non-parameter Regression Method and Kalman Filtering Method

准确预测公交车到站时间并实时向公交乘客发布,是提高公交车服务水平的关键因素。在采集的公交车GPS数据的基础上,建立了组合非参数回归和卡尔曼滤波的公交车到站时间预测模型。以北京公交300路内为例,选取1 020组公交车行驶数据,取其中128组作为验证数据,选用平均绝对误差率作为评价指标,详细评价了不同时段及滤波前后模型的预测效果。预测结果表明,该模型具有较高的预测精度。

Accurate bus arrival time prediction and real-time published will contribute to improve the bus serv- ice level. Using the GPS data collected by the bus, the thesis builds the bus arrival time prediction model based on non-parameter regression method and Kalman filtering method. And then, this thesis gives the case study of the 1 020 group data from the inner line 300 of Beijing Bus, taking 128 from them as validate data. It uses the index of Mean Absolute Percentage Error （MAPE） to evaluate the model＇ s performance on different periods and before and after filtering. The result shows that this model has high prediction accuracy.