考虑前序路段状态的公交到站时间双层BPNN 预测模型

为提高公交到站时间预测精度，提出基于双层BPNN与前序路段状态的综合预测模型. 基于静态变量及顶层BPNN模型预测车辆到达每个站点的初始行程时间，利用K-means 聚类及马尔科夫链模型基于前序路段状态预测目标路段行驶时间；将上述两个模型的预测值及上一班次车辆的行程时间作为输入变量，基于底层BPNN模型预测车辆在目标路段的行程时间，进而动态调整车辆到达每个站点的时间. 以上海市791 路公交车早晚高峰各路段的行程时间为例进行模型测试，并与其他4 种模型进行比较. 结果表明，所提模型具有较高的预测精度，尤其在雨天，比传统BPNN模型预测精度提高57.25%.

Bi-layer BPNN Prediction Model for Bus Arrival Time Considering Preceding Segment State

An integrated prediction model is proposed based on the bilevel BPNN and the state of the presequence road section in order to improve the prediction accuracy of bus arrival time. Based on the static variables and the upper-level BPNN model, the initial travel time of the vehicle to each station is predicted. The K-means clustering method and the Markov chain model with the state of the pre- sequence road section are adopted to predict the travel time at the targeted road section. Taking the predicted values of the above two models and the travel time of the previous vehicle as input variables, the travel time of the vehicle on the targeted road segment is predicted based on the underlying BPNN model, and then the arrival time of the vehicle at each station is dynamically adjusted. Taking the travel time of Bus Route 791 in Shanghai in the morning and evening peaks as an example, the model test was performed and compared with the other four models. The results indicate that the proposed model shows higher prediction accuracy. Especially on rainy days, it improves the prediction accuracy by 57.25% compared with the traditional BPNN model.