The Research And Application Of Traffic Characteristics Based On Cellular Signaling Data

At present,traffic issues have become an important topic for urban sustainable development.The global technological innovation characterized by data-driven and intelligence is bringing major changes to the transportation industry.Cellular signaling data has the advantages of full-area and fullnetwork coverage,and can analyze individual travel behavior and group traffic patterns at a low cost.Research on traffic characteristics mainly includes travel mode recognition and traffic flow prediction.Scientifically and efficiently analyzing and predicting traffic characteristics is crucial to the construction of traffic systems.This paper studies traffic characteristics based on cellular signaling data.The main work contents and contributions are as follows:Firstly,to solve the problem that existing travel mode recognition methods are limited to coarsegrained mode and do not dig into the multi-dimensional travel characteristics of signaling data,this paper proposes a travel mode recognition method based on cellular signaling data.Firstly,based on the preprocessed cellular signaling data,the features of signaling track points and track segments are extracted,and the PCA dimensionality reduction method is used for feature screening.Then,combined with K-Means clustering features and membership features,the final multi-dimensional feature data set is formed.Finally,the genetic algorithm is used to obtain the optimal parameter combination of the XGBoost model,and the multi-dimensional travel feature set is input to realize the identification of the user’s travel mode.The experimental simulation shows that the classification performance of the GA-XGBoost model is better than that of the traditional machine learning model.Compared with the basic XGBoost model,the accuracy of this model(92.5%)is increased by 3.4%.Considering the difference and influence of peak hours and non-peak hours,the classification effect of the GA-XGBoost model is the best,which proves that the model has strong robustness in different scenarios.Then,to solve the problem that existing traffic flow prediction methods do not combine other factors affecting traffic flow dynamics and only process time series information from one direction without fully mining the spatio-temporal characteristics,this paper proposes a traffic flow prediction method based on cellular signaling data.Firstly,based on the road vehicle flow data and weather information,the input matrix is constructed with multi-dimensional information.Then,the CNN model is used to extract the spatial characteristics of the traffic flow data,and the Bi LSTM network is used to capture the changing trend of traffic flow from the forward and backward directions,so as to extract the time characteristics.Finally,an attention mechanism is introduced to assign higher weights to important features to achieve accurate traffic flow prediction.The experimental simulation shows that the prediction performance of CNN-Bi LSTM-Attention model is ideal.Taking the sampling interval of 5 minutes as an example,the MAPE and RMSE of this model are 6.1% and18.53,respectively.And the traffic flow under different travel modes is input for verification,the best prediction effect is also obtained,which proves that the model has strong generalization under different data sets.Finally,since the abstract traffic data is often difficult to be understood and used by travelers,this paper designs and implements a traffic characteristics visualization system based on Flask and ECharts framework.The system integrates the analysis results of the above algorithm model,and implements it in combination with various visualization tool libraries and Baidu map API.The system mainly includes core modules such as system management,data preprocessing,travel mode and traffic flow.At the same time,it provides the display of regional heat maps,travel trajectories,travel mode identification and proportion analysis,traffic flow query and prediction and other functions,and provides technical basis and platform support for urban travel behavior analysis and traffic flow forecasting.