[1Yin-B] Poster 2

[1Yin-B-61] Feature Selection for Regression Using Associative Local Distribution Rule Mining

〇Tomonori Obana¹, Kaoru Shimada¹ (1. Gunma University)

Keywords:

Feature selection,Association Rule,Regression

Previous studies on filter-based feature selection methods have primarily focused on classification problems, while research on regression problems remains limited. In this study, we propose a new filter method utilizing the itemsets with statistically distinctive backgrounds(ItemSB) derived from local distributions for continuous target variables. This approach enables the quantification of each feature’s contribution to characterizing the local distribution of the target variable. Based on this quantification, a subset of promising features is identified. Experiments conducted on public datasets demonstrate that our approach consistently achieves performance comparable to or superior to existing methods, confirming its effectiveness. Furthermore, we also considered the explainability of proposed method.