[4Xin1] Poster session 2

[4Xin1-49] A clustering method for research project based on research style in JSPS KAKENHI Grant

研究費助成事業の審査区分に対する適用例

〇Hidekazu Nakawatase¹ (1.National Institute of Informatics)

Keywords:EDA, discipline, Research evaluation, clustering, Explainable AI

Although journal articles are not the only format for publishing research results, the number of articles and their citations often attracts attention as indicators for research evaluation. In order to evaluate research appropriately and fairly, it is necessary to use suitable indicators for each research field. In this study, we examined an analysis method that uses decision trees to cluster a set of research projects according to their publication format and also generates classification criteria. Applying this method to a group of KAKENHI projects, we were able to generate clusters corresponding to major categories such as humanities, mathematical and physical sciences, medicine, and so on. The generated classification criteria also revealed the primary publication format for each cluster.