VDS Data-Based Deep Learning Approach for Traffic Forecasting Using LSTM Network

摘要

Traffic forecasting is an important component of the Intelligent Transportation System (ITS). Recently, deep learning has been introduced as a promising method for traffic forecasting to deal with the exponential growth of data in ITS. In this regard, this paper focuses on applying a deep neural network model using LSTM for traffic forecasting based on analyzing data from the Vehicle Detection System (VDS). In particular, we first try to understand the traffic condition by applying visualization techniques. Then, based on the traffic condition, we apply an appropriate deep learning model for predicting traffic flow. Experiments in a certain urban area present promising results by applying the proposed model.