本研究では、深層学習により産業連関表の投入係数行列を推計する新たな手法を提示する。この研究では都道府県のデータを用いた深層学習を試みるが、そのデータが小さいために過学習を引き起こす可能性が高い。そこで、複数の都道府県を一つにまとめた仮想的な地域のデータを生成するデータ拡張により、データサイズを増加させ過学習の回避を試みる。山口県および岐阜県郡上市について投入係数行列を予測した結果、深層学習による推計方法はRAS法より安定した精度で投入係数を予測できうることを示した。

Abstract: This research clarifies the economic ripple effect of contents tourism implemented mainly in Ube City, Yamaguchi Prefecture. The target content is a group of works created by Hideaki Anno, a video creator from Ube City. The analysis method utilizes an input-output table. As a result, it became clear that this tourism policy has produced an economic effect of approximately 1.3 billion yen. On the other hand, because input-output analysis requires specialized knowledge, there have been few cases of using input-output analysis to verify policies promoted by local governments at the municipal level in Yamaguchi Prefecture. In the future, we will continue to analyze the economic ripple effect, especially for measures taken at the municipal level, and quantitatively verify the measures.