Study On The Optimal Time Granularity Of Short-term Passenger Flow Forecasting In Urban Metro

With the gradual formation of urban metro transit networks,urban rail transit organizations are becoming more and more complex,and the requirements for passenger flow management are becoming more intelligent,dynamic and refined.Mastering the change rule of the real-time passenger flow has become a focus of track operation,and the variation rule of passenger flow is affected by the time interval of passenger flow observation.Previous studies have not explicitly considered the influence of time granularity.Most of them stay in the analysis of passenger flow information under a single time granularity,which makes it difficult to fully exploit the effective information contained in the sequence itself,which leads to the inability to accurately describe the changing trend of similar rules among passenger flows.So how to choose a suitable time granularity is one of the directions we need to focus on metrics.Based on this problem,this paper conducts the following research on the passenger flow in different days and the same day:Initially,the mechanism of the formation of metro passenger flow is discussed.The factors affecting the distribution of passenger flow are analyzed from the internal and external aspects of the system.On this basis,the distribution rule of passenger flow at time and space is analyzed,which provides a solid theoretical basis for the study of the relationship between time granularity and passenger flow rule.Then,the optimal time granularity selection and passenger flow forecast under the different days same week and same days different week.According to the variation rules and volatility characteristics of the historical passenger flow in the same day and different days,firstly,using the combination of similar metric coefficient method and cluster analysis method,the multiple time granularities selected by the paper are classified and combined.The optimal time granularity is that it can make the inbound passenger flow show strong similarity regularity and retain the characteristics of the passenger flow itself.Then,according to the characteristics of the fluctuation of passenger flow on the same day and different day,the appropriate prediction model is constructed.Finally,the optimal time granularity selected above is used for two ways of the different week and the same week to predict the assessment day,and the feasibility of the time granularity selection of the paper is verified by the reasonableness of prediction accuracy.Finally,the correctness of time granularity selection is verified by taking the passenger flow of the Nanping site as an example.According to the case study,the optimal time granularity for short-term passenger flow forecast on the same day and the different day are 10 min.The 10 min is used as the passenger flow input time interval,and the two paths are predicted by the SARIMA model and the GM-SARIMA model respectively.The results of prediction are within the allowable error range,indicating that the prediction model can fit the passenger flow with periodic fluctuation characteristics well;in addition,the passenger flow prediction results at 1 min,3 min,5 min,7 min,and 15 min were compared with the predicted results at 10 min.The results show that the passenger flow prediction is better than other time granularities when the time granularity is 10 min.From this,we can verify the rationality and accuracy of the method of selecting the best time granularity.