[공지사항]을 빙자한 안부와 근황 
Show more

Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Towards Explainable Anomaly Detection in Shared Mobility Systems

Created by
  • Haebom

Author

Elnur Isgandarov, Matteo Cederle, Federico Chiariotti, Gian Antonio Susto

Outline

This paper presents an interpretable framework for anomaly detection in shared mobility systems such as bike sharing networks. By integrating various data sources such as bike rental records, weather conditions, and public transportation availability, unsupervised learning-based anomaly detection using the Isolation Forest algorithm is performed, and the interpretability is enhanced through the Depth-based Isolation Forest Feature Importance (DIFFI) algorithm. Through station-level analysis, the impact of external factors such as bad weather or public transportation restrictions is highlighted, and the results show that it contributes to improving shared mobility operation decision-making.

Takeaways, Limitations

•
Takeaways:
◦
Presenting an anomaly detection framework for shared mobility systems that integrates various data sources
◦
Building an interpretable anomaly detection model using the Isolation Forest and DIFFI algorithms
◦
Analysis of the impact of external factors through stop unit analysis and suggestion of directions for improving shared mobility operation
•
Limitations:
◦
Lack of detailed description of specific datasets or hyperparameter settings for algorithms.
◦
Absence of performance comparison analysis with other anomaly detection algorithms
◦
Additional research is needed on the effectiveness and limitations when applied to actual operating environments.
👍
Made with Slashpage