Presentation 2021-05-28
Proposal of Classifier Ensembles via Domain Knowledge for anomaly detection
Masafumi Tsuyuki, Soichi Takashige, Daisuke Komaki,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Anomaly detection models based on sensor data for preventive maintenance of facilities, need to be updated based on continuous evaluation (accuracy and interpretability). However, since the correct labels cannot be obtained for the anomaly detection task, we must evaluate the prediction results by visual confirmation. In this study, we propose an ensemble model that combines several models specialized for specific types of failures, which are created from domain knowledge. The proposed ensemble model is expected to reduce the amount of visual check and the side effects of updating by combining the models before and after updating, thus making the model updating process more efficient.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Machine Learning / Anomaly Detection / Ensemble / Domain Knowledge / MLOps
Paper # R2021-2
Date of Issue 2021-05-21 (R)

Conference Information
Committee R
Conference Date 2021/5/28(1days)
Place (in Japanese) (See Japanese page)
Place (in English) Virtual
Topics (in Japanese) (See Japanese page)
Topics (in English) Software Reliability, Reliability General
Chair Akira Asato(Fujitsu)
Vice Chair Tadashi Dohi(Hiroshima Univ.)
Secretary Tadashi Dohi(Hosei Univ.)
Assistant Hiroyuki Okamura(Hiroshima Univ.) / Shinji Yokogawa(Univ. of Electro-Comm.)

Paper Information
Registration To Technical Committee on Reliability
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Proposal of Classifier Ensembles via Domain Knowledge for anomaly detection
Sub Title (in English)
Keyword(1) Machine Learning
Keyword(2) Anomaly Detection
Keyword(3) Ensemble
Keyword(4) Domain Knowledge
Keyword(5) MLOps
1st Author's Name Masafumi Tsuyuki
1st Author's Affiliation Hitachi, Ltd.(Hitachi)
2nd Author's Name Soichi Takashige
2nd Author's Affiliation Hitachi, Ltd.(Hitachi)
3rd Author's Name Daisuke Komaki
3rd Author's Affiliation Hitachi, Ltd.(Hitachi)
Date 2021-05-28
Paper # R2021-2
Volume (vol) vol.121
Number (no) R-47
Page pp.pp.7-12(R),
#Pages 6
Date of Issue 2021-05-21 (R)