Long short-term memory model for traffic congestion prediction with online open data

Traffic congestion in metropolitan areas has become more and more serious. Over the past decades, many academic and industrial efforts have been made to alleviate this problem, among which providing accurate, timely and predictive traffic conditions is a promising approach. Nowadays, online open data have rich traffic related information. Typical such resources include official websites of traffic management and operations, web-based map services (like Google map), weather forecasting websites, and local events (sport games, music concerts, etc.) websites. In this paper, online open data are discussed to provide traffic related information. Traffic conditions collected from web based map services are used to demonstrate the feasibility. The stacked long short-term memory model, a kind of deep architecture, is used to learn and predict the patterns of traffic conditions. Experimental results show that the proposed model for traffic condition prediction has superior performance over multilayer perceptron model, decision tree model and support vector machine model.