Dense velocity prediction via Deep learning using ETC2.0 probe data

Abstract

In order to predict highway travel time, we performed machine learning using vehicle detector data in a previous study. But the data of the vehicle detector is about 2 [km] apart, and the length of the traffic congestion cannot be accurately expressed. Therefore, it is expected to improve the accuracy by using ETC2.0 probe data which can obtain more detailed data. However, detailed data does not always improve the accuracy, and some problems are expected. In this paper, we will perform machine learning using ETC2.0 probe data, and verify its characteristics by comparing it with that of vehicle detector data.