Traffic State Prediction using Convolutional Neural Network

Traffic state prediction methods have been considered by many researchers since accurate traffic prediction is an important part of the successful implementation of the Intelligent Transportation System (ITS). This study develops the traffic prediction model based on real traffic data in congested hours of expressways in Bangkok, Thailand. Unlike most studies, this model utilizes data from 40 nodes along the expressway instead of a single sensor. A Convolutional Neural Network (CNN) model was applied and compared to other widely used models. The result shows that the accuracy of CNN model is higher than other models.