講演抄録/キーワード |
講演名 |
2015-08-05 13:00
Empirical Study on Instance Selection from Healthcare Data Using Fuzzy Rough Sets ○Do Van Nguyen(KDDI R&D Labs)・Keisuke Ogawa(KDDI)・Kazunori Matsumoto・Masayuki Hashimoto(KDDI R&D Labs) DE2015-14 |
抄録 |
(和) |
This research shows a study on issues that cause low performance in disease prediction such as overlap and imbalance and suggests solutions based on fuzzy rough sets to overcome these problems. We apply several methods based on rough set theory to deal with overlap by selecting valuable instances. The considerable results in combination of rough instance selection and balancing technique are then noticed. In this study, we also apply new proposed approach that uses only fuzzy-rough instance selection to dealing with the both issues. This means that without balancing techniques, performance still improves. Some experiments are conducted to show the methods. |
(英) |
This research shows a study on issues that cause low performance in disease prediction such as overlap and imbalance and suggests solutions based on fuzzy rough sets to overcome these problems. We apply several methods based on rough set theory to deal with overlap by selecting valuable instances. The considerable results in combination of rough instance selection and balancing technique are then noticed. In this study, we also apply new proposed approach that uses only fuzzy-rough instance selection to dealing with the both issues. This means that without balancing techniques, performance still improves. Some experiments are conducted to show the methods. |
キーワード |
(和) |
rough set theory / fuzzy-rough sets / healthcare data / classification performance / instance selection / / / |
(英) |
rough set theory / fuzzy-rough sets / healthcare data / classification performance / instance selection / / / |
文献情報 |
信学技報, vol. 115, no. 177, DE2015-14, pp. 19-24, 2015年8月. |
資料番号 |
DE2015-14 |
発行日 |
2015-07-29 (DE) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
DE2015-14 |