Machine Learning-based traffic prediction models for Intelligent Transportation Systems

Abstract

Intelligent Transportation Systems (ITS) have attracted an increasing amount of attention in recent years. Thanks to the fast development of vehicular computing hardware, vehicular sensors and citywide infrastructures, many impressive applications have been proposed under the topic of ITS, such as Vehicular Cloud (VC), intelligent traffic controls, etc. These applications can bring us a safer, more efficient, and also more enjoyable transportation environment. However, an accurate and efficient traffic flow prediction system is needed to achieve these applications, which creates an opportunity for applications under ITS to deal with the possible road situation in advance. To achieve better traffic flow prediction performance, many prediction methods have been proposed, such as mathematical modeling methods, parametric methods, and non-parametric methods. Among the non-parametric methods, the one of the most famous methods today is the Machine Learning-based (ML) method. It needs less prior knowledge about the relationship among different traffic patterns, less restriction on prediction tasks, and can better fit non-linear features in traffic data. There are several sub-classes under the ML method, such as regression model, kernel-based model, etc. For all these models, it is of vital importance that we choose an appropriate type of ML model before building up a prediction system. To do this, we should have a clear view of different ML methods; we investigate not only the accuracy of different models, but the applicable scenario and sometimes the specific type of problem the model was designed for. Therefore, in this paper, we are trying to build up a clear and thorough review of different ML models, and analyze the advantages and disadvantages of these ML models. In order to do this, different ML models will be categorized based on the ML theory they use. In each category, we will first give a short introduction of the ML theory they use, and we will focus on the specific changes made to the model when applied to different prediction problems. Meanwhile, we will also compare among different categories, which will help us to have a macro overview of what types of ML methods are good at what types of prediction tasks according to their unique model features. Furthermore, we review the useful add-ons used in traffic prediction, and last but not least, we discuss the open challenges in the traffic prediction field.