基于PSO-FSVR的城市轨道交通客流预测模型

Urban Rail Transit Passenger Flow Prediction Model Based on PSO-FSVR

目的:为了提高城市轨道交通(以下简称“城轨”)运营决策效率,控制运营成本,减少资源空置和浪费,有必要对城市轨道交通客流进行精准预测。方法:提出一种基于PSO(粒子群优化)-FSVR(模糊支持向量回归机)的城轨客流预测模型。对城轨客流数据进行清洗和分析,根据获取的数据构建特征空间,利用PSO算法对FSVR模型中的参数进行优化以提高预测精度,引入模糊隶属度构建PSO-FSVR模型,增强客流预测方法的泛化能力。将该模型应用于北京地铁2号线某车站的进出站客流预测中,并引入BP(反向传播)神经网络模型与GRNN(广义回归神经网络)模型作为对比方法,采用均方根误差、显著性参数及相关系数等指标对预测结果进行评价。结果及结论:基于PSO-FSVR的城轨客流预测模型下,工作日、节假日客流预测绝对均方根误差分别为7.010 4和8.778 5,相关系数分别为0.993 0和0.955 8,相较其他两种模型,该模型客流预测性能最优。基于PSO-FSVR的城轨客流预测模型能够有效提高城轨客流预测的准确性,尤其在客流极值点处的表现比其他方法更加优秀,具有良好的预测性能。

Objective:In order to enhance the efficiency of operational decision-making,control operating costs,and reduce resource wastage in URT(urban rail transit),it is essential to accurately predict passenger flow.Method:An URT passenger flow prediction model based on PSO(particle swarm optimization)and FSVR(fuzzy support vector regression)is proposed.The passenger flow data of URT are cleaned and analyzed,and a feature space is constructed based on the acquired data.The PSO algorithm is utilized to optimize the parameters of the FSVR model for higher prediction accuracy.Fuzzy membership degrees are introduced to construct the PSO-FSVR model,thereby improving the model generalization ability.This model is applied to predict the inbound and outbound passenger flow of a station on Beijing Subway Line 2.BP(backpropagation)neural network and GRNN(general regression neural network)models are taken as comparative methods.The evaluation of prediction results is carried out using metrics such as RMSE(root mean square error),significance parameters,and correlation coefficients.Result&Conclusion:Under the PSO-FSVR-based URT passenger flow prediction model,the absolute RMSE for weekday and holiday passenger flow predictions are 7.0104 and 8.7785,respectively;the corresponding correlation coefficients are 0.9930 and 0.9558,respectively.Compared to the other two models,this model exhibits the best passenger flow prediction performance.The PSO-FSVR-based URT passenger flow prediction model effectively enhances the accuracy of passenger flow prediction,particularly excelling at predicting passenger flow extremes,demonstrating excellent predictive performance.