基于PSO-LightGBM的公交行程时间预测

Bus Arrival Time Prediction Based on PSO-LightGBM Algorithm

为有效提升城市公交到站预测的准确性,提高城市智能公交系统服务水平,提出了1种基于粒子群优化算法(PSO)和轻量级梯度提升机(LightGBM)的公交行程时间预测模型,采用LightGBM构建公交行程时间的预测模型,PSO用于优化LightGBM模型的超参数.以广州市某典型公交线路为例,从公交调度系统中提取大量的历史数据进行预处理,分析公交行程时间的影响因素,构建模型特征集,建立了基于PSO-LightGBM、LightGBM、BP神经网络和LSTM的公交行程时间预测模型.结果表明PSO-LightGBM模型在早高峰、晚高峰和平峰时段的平均绝对百分比误差(MAPE)分别为7.72%、7.56%和7.31%,该模型对公交行程时间的预测准确率要高于LightGBM模型、BP神经网络模型和LSTM模型.

In order to improve the prediction accuracy of bus arrival time and increase levels of public transport system service,a new prediction model for the bus arrival time based on particle swarm optimization(PSO)and light gradient boosting machine(LightGBM)is proposed in this paper.The LightGBM model is used to predict the bus arrival time,and PSO is used to optimize the parameters during the LightGBM training.A typical bus route of Guangzhou is selected as a case study.The raw data is extracted and preprocessed from the bus scheduling system,and then the influencing factors of bus arrival time are analyzed.The feature set is constructed,then the bus arrival time prediction models are established and compared based on PSO-LightGBM,LightGBM,BP network and LSTM.The results demonstrate that the performance of PSO-LightGBM is much better than that of the others,where the mean absolute percentage error(MAPE)of PSO-LightGBM during morning peak,evening peak and non-peak are 7.72%,7.56%and 7.31%,respectively.