Highly Interpretable Agitation Detection Model by Reducing Features with Permutation Importance
This paper proposes a highly interpretable agitation detection model through reduction of features with permutation importance. We show that the proposed method can reduce the number of features from more than 500 to 15 while maintaining the accuracy of the prior agitation detection.
Authors
- Kosuke Nishihara
- Eiji Yumoto (ORCID: https://orcid.org/0000-0003-1492-3415)
- Masahiro Hayashitani
Institutions
- NEC (Japan) (JP)
Publication Details
- Published
- 2024-11-18
- DOI
- https://doi.org/10.1109/healthcom60970.2024.10880767
- Primary Topic
- Network Security and Intrusion Detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00