基于 Prophet 算法的铁路客流量预测研究-《计算机技术与发展》

文章信息/Info

Title:: Research on Railway Passenger Flow Forecast Based on Prophet Time Series Algorithm

Author(s):: WANG Xiao; CHUAI Jin-hua; ZHANG Li-heng; School of Information Engineering,Chang’an University,Xi’an 710064,China

Keywords:: dedicated passenger railway line; traffic; time series prediction; Prophet algorithm; holiday effect

摘要:: 客流量预测是铁路部门了解日常客运流量和冷热门线路具体情况的基础,是制定运输方案和列车开行计划的重要依据。针对铁路客流量变化受到季节周期、突发事件和节假日等因素影响的现象,提出一种基于 Prophet 时间序列算法的铁路客流量预测研究的新方法。实验选取了某铁路客运专线 2015 年至 2016 年的日客流量数据,经过预处理的客流量数据进行标准化得到客流量时间序列;构建节假日特征时间窗口列表和设置趋势的突变点,结合 Prophet 算法对节假日特征和突变点进行计算,并依此建立 Prophet 预测模型;利用可视化技术分析预测结果,调节参数进一步优化模型。重点结合节假日效应对未来两周客流量进行了预测分析。结果表明,Prophet 模型预测结果精确度高于 LSTM 模型,所建立的模型预测结果是合理和可靠的。

Abstract:: The passenger flow forecast is the basis for railway departments to understand the daily passenger flow and the specific situation of cold and hot lines,and it is also an important basis for formulating transport plans and train running plans. A new method for railway passenger flow prediction based on Prophet time series algorithm is proposed to solve the problem that railway passenger flow is affected by seasonal cycles,emergencies and holidays. In the experiment, daily passenger flow data of a railway passenger dedicated line from 2015 to 2016 were selected. We construct the holiday characteristic time window list and set up the sudden change point of trend,calculate the holiday characteristic and sudden change point with Prophet algorithm,and build the Prophet prediction model based on it.Visualization technology is used to analyze the prediction results and adjust parameters to further optimize the model. The forecast and analysis of the passenger flow in the next two weeks are made based on the holiday effect. The results show that the Prophet model is more accurate than the LSTM model,and the predicted results are reasonable and reliable.