Development of a new data summary and learning method using Directional Neighborhood Rough Set

Reference No. 2025a039
Type/Category Grant for Supporting the Advancement of Female Researchers- Short-term Visiting Researcher
Title of Research Project Development of a new data summary and learning method using Directional Neighborhood Rough Set
Principal Investigator Yoshie Ishii(Graduate School of Engineering,Kyoto University・Assistant professor)
Research Period August 4,2025. - August 8,2025.
February 16,2026. - February 20,2026.
Keyword(s) of Research Fields Rough set theory, Remote sensing
Abstract for Research Report The objective of this research is to develop an accurate and white box classifier by clarifying the features of the Directional Neighborhood Rough Set (DNRS). DNRS is one of the set theories and is suitable for solving classification problems of the data with quantitative variables for explanatory variables and categorical variables for objective variables. Although there are a lot of black box classifiers recently, rough set theory has white box features in that it can hold all derived rules. DNRS itself was just proposed as one of the machine learnings in 2022, and the mathematical nature of DNRS has not been clarified yet. Therefore, we try to mathematically clarify the nature of DNRS, especially about the general concepts in rough set theory, such as approximation, reduction, etc. In addition, we will develop algorithms to apply mathematical natures to real-world data. For instance, it is considered that the concept of reduction in rough set theory can be extended to dimensional reduction. In order to demonstrate the effectiveness of the algorithm using DNRS, we target satellite images. Satellite images have various needs such as monitoring urbanization, grasping the situation under natural disasters, and analyzing climate change, etc. For these social needs, a method that can treat a huge amount of satellite data is expected, and the proposed method in this study is considered to be suitable for them. The result expected from this study is the proposition of a new learning and summarizing method using DNRS, which can treat data with white box and effectively.
Organizing Committee Members (Workshop)
Participants (Short-term Joint Usage)
Yujie Gu(Faculty of Information Science and Electrical Engineering, Kyushu University・Associate Professor)