Short-Term Traffic Forecast of Urban Bus Stations Based on Long Short-Term Memory

Abstract

Smart Transportation is an important part of Smart City. Buses, as one of the most important modes of travel in urban public transport, not only facilitate the work and life of urban residents but also provide effective solutions to urban energy conservation and environmental protection. Improving the accuracy of passenger traffic forecast at bus stations has been a focus of research in smart public transport. To compensate for the limitations of single-station forecasting and short-term memory in traditional time series models (such as ARMA and SVR) and improve the short-term prediction accuracy of urban bus station traffic flow, we propose a long short-term memory (LSTM)-based neural network model than can learn long-term series data at multiple sites and forecast the traffic flow of multiple sites simultaneously. Experimental results show that learning multi-site traffic simultaneously can improve the accuracy of prediction results and minimize mean squared error and mean absolute error to 3.18 and 1.43, respectively. Aside from its long-term memory, the neural network model based on LSTM can also show the potential correlation between multiple sites, suggesting obvious advantages in short-term traffic forecast. This study proves the feasibility of using LSTM for multi-site bus station traffic forecast. In addition, the performance of the proposed model is considerably better than that of single-site traffic forecast. Traffic monitoring data analysis indicates that multi-site bus station traffic flow is correlated. This paper provides a timely and accurate data reference for the rapid decision-making and integrated management of public transport operations.