한국해군과학기술학회
[ Article ]
Journal of the KNST - Vol. 6, No. 3, pp.293-297
ISSN: 2635-4926 (Print)
Print publication date 30 Sep 2023
Received 18 Jul 2023 Revised 01 Aug 2023 Accepted 31 Aug 2023
DOI: https://doi.org/10.31818/JKNST.2023.09.6.3.293

설명 가능한 인공지능 (XAI)을 이용한 확률 임베딩 벡터 예측에 관한 연구

양성실1 ; 설현주2, *
1충남대학교 군사학과 박사과정
2충남대학교 국가안보융합학부 교수
Prediction of Probability Embedding Vector Using eXplainable Artificial Intelligence(XAI)
Seong-Sil Yang1 ; Hyeon-Ju Seol2, *
1Ph.D. student, Department of Military Studies, Chungnam National University
2Professor, School of Integrated National Security, Chungnam National University

Correspondence to: *Hyeon-Ju Seol School of Integrated National Security, Chungnam National University Room 216, College of Natural Sciences Bldg. II, 99 Daehak-ro, Yuseong-gu, Daejeon 34134, Republic of Korea Tel: +82-42-821-8531 Fax: +82-42-821-8531 E-mail: hjseol@cnu.ac.kr

© 2023 Korea Society for Naval Science & Technology

초록

설명 가능한 인공지능(XAI)은 기존 머신러닝의 성과와 설명력을 한 단계 더 향상시킬 수 있는 기술로 최근 주목받고 있다. 본 연구에서는 이 XAI 방법을 전훈분석이라는 국방정책에 정량적으로 새롭게 적용하고자 연구를 진행하였다. 특히 인공지능 자연어처리를 위한 LDA 토픽모델링을 적용한 새로운 전훈분석 방법을 제시한다. 이를 통해 도출된 정보를 예측하고 그 근거를 분석할 뿐만 아니라 인공지능 비전문가와 전투현장 지휘관의 신속한 판단에 도움이 되고자 한다.

Abstract

eXplanable Artificial Intelligence(XAI) has recently gained attention as a technology that can enhance the performance and explanatory power of existing machine learning systems. In this study, we conducted research to quantitatively apply XAI methods to a defense policy known as Lessons Learned(LL) Analysis. Specifically, we propose a new LL analysis method that applies LDA topic modeling for AI NLP. The objective is not only to predict the information obtained from this method and analyze its basis, but also to assist non-expert AI users and combat field commanders in making quick judgments.

Keywords:

Lessons Learned Analysis, DOTMLPF-P, Artificial Intelligence, LDA Topic Modeling, XAI

키워드:

전쟁교훈분석, 전투발전체계 분야, 인공지능, LDA 토픽 모델링, 설명 가능한 인공지능

Acknowledgments

이 논문은 2023년도 한국해군과학기술학회 하계학술대회 발표 논문임.

References

  • Gunning, David and Aha, David W., “DARPA’s Explainable Artificial Intelligence Program,” AI Magazine, Association for the Advancement of Artificial Intelligence, 2019. pp. 44-58. [https://doi.org/10.1609/aimag.v40i2.2850]
  • Yang, S-S. and Soel, H-J., “A Study on the Lessons Learned Analysis Using Artificial Intelligence Technique: Based on LDA Topic Modeling,” Review of Korean Military Studies, Vol. 12, No. 1, 2023, pp. 29-48.
  • Blei, David M., “Probabilistic Topic Models,” Communications of the ACM, Vol. 55, No. 4, 2012, pp. 77-84. [https://doi.org/10.1145/2133806.2133826]
  • Lundberg, Scott M., Erion, Gabriel G. and Lee, Su-In, “Consistent Individualized Feature Attribution for Tree Ensembles,” arXiv preprint arXiv:1802.03888, , 2018.