Proceedings of the Annual Conference of JSAI
Online ISSN : 2758-7347
37th (2023)
Session ID : 2P1-OS-23-02
Conference information

Cause Explanation based on Machine Learning Interpretability for Anomaly Detection on Demographic Data
*Ryo KOYAMATomohiro MIMURAShin ISHIGUROKeisuke KIRITOSHITakashi SUZUKIAkira YAMADA
Author information
CONFERENCE PROCEEDINGS FREE ACCESS

Details
Abstract

When accidents, disasters, or other unusual events occur, traffic is disrupted, causing congestion and difficulty in movement. In order to mitigate such situations, it is necessary to accurately detect the cause of the abnormality from human movement data and take prompt action. As a method for detecting anomalies, a method to obtain reconstruction errors by dimensionality reduction has been proposed. However, the reconstruction error obtained by this method is calculated under the influence of correlations between features, so it cannot fully explain the cause of the abnormality. Therefore, this paper calculates the SHAP values of the reconstruction error of dimensionality reduction. To verify the effectiveness of the method, a dataset of human movements was created and the proposed method was applied. The experimental results show that the proposed method can explain anomalies with higher accuracy than conventional methods.

Content from these authors
© 2023 The Japanese Society for Artificial Intelligence
Previous article Next article
feedback
Top