| 作成者 |
|
|
|
| 本文言語 |
|
| 出版者 |
|
|
|
| 発行日 |
|
| 収録物名 |
|
| 巻 |
|
| 号 |
|
| 開始ページ |
|
| 終了ページ |
|
| 出版タイプ |
|
| アクセス権 |
|
| Crossref DOI |
|
| 概要 |
Screening is a useful method for selecting important variables for high-dimensional data where the number of predictors is much larger than the sample size. Screening methods can eliminate unnecessary... variables at a low computational cost by calculating their importance scores, such as the correlation between response and predictor variables. In this study, we consider the problem of selecting variables and interactions in classification problems for data with imbalanced sample sizes between classes. Specifically, we propose a new method called Class-to-Class KIF (CCKIF) to select interactions in imbalanced multi-class classification problems. CCKIF uses the difference in Kendall’s rank correlations for each class to calculate the importance scores of the interactions to improve the selection accuracy more than existing methods, even for imbalanced data. We present the theoretical properties of the proposed method. Simulation studies and real data analysis show that the proposed CCKIF appropriately selects important interactions, particularly for data on minor classes.続きを見る
|
Mendeley出力
