Abstract
To keep road maps and map data in car navigation system current gives great benefits to citizens. However, a lot of time and effort is needed to gather information which follows roads change. To improve the situation, this paper presents construction-records which are collected to control the quality of public works all over Japan can predict for locations where roads change. First, construction-records within a certain period of time are grouped by location. Next, the construction method/ type information is converted into explanatory variables (independent variable). The variables aggregated by each location are calculated by a supervised machine learning. An experimental evaluation showed that the method outputted decision tree models which reflected process of road construction, and the average recall ratio was approximately 95%.
