2017 年 31 巻 p. 271-277
In order to conduct regional disaster prevention activities continuously and appropriately, it is necessary to evaluate the effectiveness and appropriateness, and make the improvements based on evaluation. However, there are insufficient human resources of evaluators and it is difficult to foster the talent, since extensive field experience and broad disaster prevention knowledge is required for appropriate evaluation. This study constructed a prototype of a machine learning system for automatically evaluating regional disaster prevention activities. It also performs the machine learning with the data from activity records of the Bosai Contest as input variables, and the winning judgment data evaluated by the expert's review committee as output variables. As a result, the same result as experts’ judgment was obtained with the probability of 94% for learning data and 79% for verification data.