Research On Prediction Model Of Railway Passenger Traffic Volume Based On Spatial And Temporal Characteristics Analysis

In the field of transportation in China,railway transportation has always played an important role.With the rapid development of high-speed railway construction,the road network of eight horizontal and eight longitudinal high-speed railway is further improved,which promotes the rapid growth of railway passenger transport.The prediction of the future passenger traffic volume of railway is helpful to the reasonable operation and management planning of railway,and thus promote the healthy and stable development of railway traffic.At present,the traditional statistical prediction method is mainly used in the railway passenger transport forecasting algorithm,and few researches on the field of deep learning are carried out.The research on deep learning algorithm mainly focuses on mining the historical and temporal rules of data,but neglecting the spatial correlation of each province.At the same time,the data characteristics of railway passenger transport and economy are closely related,and the influencing factors are multiple and complex.This paper studies the prediction model of railway passenger transport based on timespace prediction,and the specific research contents are as follows:(1)In view of the numerous factors affecting railway passenger transport,the existing influencing factors are complex in relation to each other,and the influencing factors are quantifiable and quantifiable.Through the full analysis of the relationship between the influencing factors and the passenger traffic volume,the influencing factors are analyzed and selected for the input of the later order prediction model.Firstly,qualitative analysis is used to establish the index system of influencing factors,and then the correlation between the influencing factors and passenger traffic volume is calculated by quantitative analysis.The feature selection algorithm based on correlation degree is used to filter the features.The feature subset selected finally meets the requirements of low correlation degree and high correlation degree with passenger traffic volume.Finally,the validity of feature selection is verified on the data set of railway passenger transport.(2)Aiming at the problem that the passenger transport forecast of railway is not only related to the historical time series changes,but also the passenger volume of each province has certain correlation in space,a multi factor spatiotemporal prediction model based on multi graph convolution is proposed.First,the spatial relationship of each province is fully excavated,and the spatial relationship between provinces is obtained by constructing multiple spatial feature maps.Then,the temporal and spatial features are fused by volume product.At the same time,the influence factors selected by the analysis are fused by multi-layer perceptron and graph convolution results,so as to obtain the final prediction results and improve the accuracy of the prediction model.Finally,the algorithm is analyzed and verified on the data set of railway passenger transport.(3)Aiming at the problems of scattered data and large amount of query,a set of railway passenger transport visualization platform is developed for manual report generation.The system uses Vue and spring boot to realize the separation of front and rear ends,shows the data analysis results of railway passenger transport,and integrates the prediction model algorithm of railway passenger transport.The platform uses the means of line chart,histogram,table and other means to visualize the data.