Developing an Input-Output Table Generation Algorithm Using a Japanese Trade Database: Dealing with Ambiguous Export and Import Information

In this study, we undertook the task of data conversion of enterprise export/import ratio using text mining. In addition, an input-output table was generated based on corporate data and information by using an algorithm. We also address the lack of input-output tables pertaining to individual companies in the reports released by various government ministries and agencies. Moreover, sales information and cost information of individual companies include the sales obtained through overseas exports and costs due to overseas imports. Therefore, in order to accurately calculate what is produced in one country, it is necessary to distinguish and separate the sales revenues and costs incurred from overseas trade. In this work, we structured the data. Text mining was applied to the information described by free descriptions of import/export ratios of direct trade companies to sales. Then, we simulate using structured data and propose an algorithm to construct an input-output table that includes foreign transactions.