[1I4-GS-2] Machine learning: Applied machine learning (1)

[1I4-GS-2-05] Feature Analysis of Road Network and its Generation

〇Genya Nobuhara¹, Hideaki Uchida¹, Kazuki Abe¹, Hideki Fujii¹, Shinobu Yoshimura¹ (1. the University of Tokyo)

Keywords:road network, feature value, L-system, machine learning

When carrying out a social system simulation (e.g. traffic simulation) using geographic information, there are cases to be convenient to use a virtual road network having characteristics close to those of a real road network. The purpose of this study is to create a virtual road network that reproduces the features of the real road network. Real road network data was obtained from OpenStreetMap, an open-source geographic information database. The features of the road network were calculated from the acquired data, and the virtual road network having similar features was generated by the L-system. Input parameters of L-system were determined by regression using machine learning. As a result of experiments using road networks of multiple model cities, it was confirmed that a virtual road network that reproduces many features of the real road network could be generated by properly setting the input parameters of the L-system.