Traffic Flow Prediction with Swiss Open Data: A Deep Learning Approach

Open government data (OGD) are provided by the public sector and governments in an open, freely accessible format. Among various types of OGD, dynamic data generated by sensors, such as traffic data, can be utilized to develop innovative artificial intelligence (AI) algorithms and applications. As AI algorithms, specifically Deep Neural Networks, necessitate large amounts of data, dynamic OGD datasets serve as supplemental resources to existing traffic datasets, used for performance comparison and benchmarking. This work examines the effectiveness of using open traffic data from the Swiss open data portal to develop a Graph Neural Network (GNN) model for traffic forecasting. To this end, the objective of this study is to probe the extent to which dynamic OGD can enhance the accuracy and efficiency of traffic forecasting models, and more critically, to investigate the potential of this data in driving the development of cutting-edge AI models for traffic flow prediction. We posit that strategic utilization of such data has the potential to catalyze a transformative shift in the realm of traffic management and control, by fostering intelligent solutions that effectively leverage the predictive capabilities of AI models. The results indicate that the GNN-based algorithm is effective in predicting future traffic flow, outperforming two traditional baselines for time series forecasting.