【作者】 李鹏程； 孙敏； 庞溟舟；

【Author】 Pengcheng Li;Min Sun;Mingzhou Pang;School of Electrical Information and Electrical Engineering,Shanghai Jiao Tong University;

【机构】 上海交通大学电子信息与电气工程学院；

【摘要】 准确的公交到站时间预测对提高乘客的出行效率以及改善公交系统的服务水平具有重要的意义。相对于传统的预测方法,本文提出了一种基于长短时记忆(long short-term memory,LSTM)神经网络的公交到站时间预测方法,包括网络模型结构的设计,参数的选择,数据的处理以及网络的训练和预测,并用相关指标对模型的预测性能进行评价。此外,还就单层LSTM模型和两层LSTM模型下网络的预测情况进行了比较与分析。最后用实测数据进行验证,结果表明,文中方法能够有效地预测公交到站时间,预测精度较高。更多还原

【Abstract】 The accurate prediction of bus arrival time is of great importance to the improvement on the travel efficiency for passengers,as well as to the enhancement of public transport system services.Compared with traditional prediction models,this paper presents a prediction model for the bus arrival time forecasting based on long short-term memory(LSTM) neural network,including the design of the structure of the network model,the search of optimal parameter values,data processing and the model training and prediction.Some relevant criteria will be used to evaluate the prediction performance of this model.Besides,the paper also compares and analyses the prediction results due to the single-layer LSTM model and the double-layers LSTM model.The model is tested by the real data and the results show that the prediction model proposed in this paper has a good performance in forecasting the bus arrival time with a favorable precision.更多还原