2024 Volume 80 Issue 20 Article ID: 24-20036
In recent years, traffic flow analysis using AI cameras and micro traffic simulation has been progressing. In this study, we developed a machine learning model that enables real-time traffic jam prediction using edge AI, which discriminates vehicles at observed locations. A machine learning model that approximates a micro traffic simulation was constructed for the Shinkawa Interchange in Sapporo, and the effect of vehicle discrimination accuracy of edge AI on traffic jam prediction was evaluated. LGBM outperformed NN when there was variation. It was also confirmed that using machine learning models as an alternative to micro traffic simulation can significantly reduce the forecasting time. Further improvement of accuracy will be an issue in the future.
