Evaluating Explanation Methods for Multivariate Time Series Classification

被引:2
|
作者
Serramazza, Davide Italo [1 ]
Nguyen, Thu Trang [1 ]
Le Nguyen, Thach [1 ]
Ifrim, Georgiana [1 ]
机构
[1] Univ Coll Dublin, Sch Comp Sci, Dublin, Ireland
基金
爱尔兰科学基金会;
关键词
Time Series Classification; Explanation; Evaluation;
D O I
10.1007/978-3-031-49896-1_11
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Multivariate time series classification is an important computational task arising in applications where data is recorded over time and over multiple channels. For example, a smartwatch can record the acceleration and orientation of a person's motion, and these signals are recorded as multivariate time series. We can classify this data to understand and predict human movement and various properties such as fitness levels. In many applications classification alone is not enough, we often need to classify but also understand what the model learns (e.g., why was a prediction given, based on what information in the data). The main focus of this paper is on analysing and evaluating explanation methods tailored to Multivariate Time Series Classification (MTSC). We focus on saliency-based explanation methods that can point out the most relevant channels and time series points for the classification decision. We analyse two popular and accurate multivariate time series classifiers, ROCKET and dResNet, as well as two popular explanation methods, SHAP and dCAM. We study these methods on 3 synthetic datasets and 2 real-world datasets and provide a quantitative and qualitative analysis of the explanations provided. We find that flattening the multivariate datasets by concatenating the channels works as well as using multivariate classifiers directly and adaptations of SHAP for MTSC work quite well. Additionally, we also find that the popular synthetic datasets we used are not suitable for time series analysis.
引用
收藏
页码:159 / 175
页数:17
相关论文
共 50 条
  • [1] Improving the Evaluation and Actionability of Explanation Methods for Multivariate Time Series Classification
    Serramazza, Davide Italo
    Thach Le Nguyen
    Ifrim, Georgiana
    MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES: RESEARCH TRACK, PT IV, ECML PKDD 2024, 2024, 14944 : 177 - 195
  • [2] An Explanation Module for Deep Neural Networks Facing Multivariate Time Series Classification
    Yang, Chao
    Wang, Xianzhi
    Yao, Lina
    Jiang, Jing
    Xu, Guandong
    AI 2021: ADVANCES IN ARTIFICIAL INTELLIGENCE, 2022, 13151 : 3 - 14
  • [3] Stacking for multivariate time series classification
    Oscar J. Prieto
    Carlos J. Alonso-González
    Juan J. Rodríguez
    Pattern Analysis and Applications, 2015, 18 : 297 - 312
  • [4] Early classification on multivariate time series
    He, Guoliang
    Duan, Yong
    Peng, Rong
    Jing, Xiaoyuan
    Qian, Tieyun
    Wang, Lingling
    NEUROCOMPUTING, 2015, 149 : 777 - 787
  • [5] Stacking for multivariate time series classification
    Prieto, Oscar J.
    Alonso-Gonzalez, Carlos J.
    Rodriguez, Juan J.
    PATTERN ANALYSIS AND APPLICATIONS, 2015, 18 (02) : 297 - 312
  • [6] Agnostic local explanation for time series classification
    Guilleme, Mael
    Masson, Veronique
    Roze, Laurence
    Termier, Alexandre
    2019 IEEE 31ST INTERNATIONAL CONFERENCE ON TOOLS WITH ARTIFICIAL INTELLIGENCE (ICTAI 2019), 2019, : 432 - 439
  • [7] A short tutorial for multivariate time series explanation using tsCaptum
    Serramazza, Davide Italo
    Le Nguyen, Thach
    Ifrim, Georgiana
    SOFTWARE IMPACTS, 2024, 22
  • [8] Comparison and classification of stationary multivariate time series
    Dept. of Economet. and Bus. Stat., Monash Univ. - Caulfield Campus, P.O. Box 197 Caulfield East, Melbourne, Vic. 3145, Australia
    Pattern Recogn., 7 (1129-1138):
  • [9] A Proximity Forest for Multivariate Time Series Classification
    Zhang, Yue
    Wang, Zhihai
    Yuan, Jidong
    ADVANCES IN KNOWLEDGE DISCOVERY AND DATA MINING, PAKDD 2021, PT I, 2021, 12712 : 766 - 778
  • [10] Multilabel Classification With Multivariate Time Series Predictors
    Che, Yuezhang
    Zhu, Yunzhang
    Shen, Xiaotong
    IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2020, 68 : 5696 - 5705