Research on Traffic Flow Forecast Based on Cellular Signaling Data

Abstract

Nowadays, with the continuous increase in smartphones, a large amount of cellular signaling data is acquired by telecom operators. Research and analysis of these data can further provide help for travel mode classification and urban traffic flow prediction research. Traditional traffic flow prediction research usually uses open data sets or global positioning system (GPS) data as the data source, only uses traffic flow data as the input of the prediction model during the prediction process, ignoring the impact of travel modes and time information on the traffic flow. In addition, GPS data collection methods are limited, and the acquisition costs are high. In order to solve the above problems, this paper designs a travel mode classification method based on cellular signaling data and uses this method to obtain travel mode characteristics, then convert the time information in the data set into time characteristics. Finally, the travel mode characteristics, time characteristics, and traffic flow data are input into the long short-term memory network (LSTM) model for traffic flow prediction. Experimental results show that by combining traffic flow data with travel mode characteristics and time characteristics, the prediction model can achieve better performance. When the input characteristics are consistent, the LSTM model has the best prediction effect than the traditional traffic flow prediction models.