Urban Traffic Dynamic OD Prediction Based on Multi-Source Data

Urban traffic OD prediction has always been a hot research topic in the field of transportation. However, most of the existing OD prediction researches are under normal conditions, without considering the influence of holidays, temperatures, weather, and other factors. This paper proposes an urban traffic OD prediction model based on multi-source data. Firstly, traffic modes are divided based on travel trajectory, speed, acceleration and other factors, and OD data within a certain time granularity are extracted. The OD pairs integrating multiple factors is predicted based on long short-term memory (LSTM) networks. By comparing with the model without using multi-source data, the results show that the LSTM model with multiple factors has higher prediction accuracy and is a better prediction method.