Daily Arxiv

This page organizes papers related to artificial intelligence published around the world.
This page is summarized using Google Gemini and is operated on a non-profit basis.
The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.

Time-IMM: A Dataset and Benchmark for Irregular Multimodal Multivariate Time Series

Created by
  • Haebom

Author

Ching Chang, Jeehyun Hwang, Yidan Shi, Haixin Wang, Wen-Chih Peng, Tien-Fu Chen, Wei Wang

Outline

This paper introduces the Time-IMM dataset and the IMM-TSF benchmark library to address time series data with irregularities, multimodality, and missing values encountered in real-world applications such as healthcare, climate modeling, and finance. Time-IMM provides data representing nine types of time series irregularities categorized by trigger-based, constraint-based, and artifact-based mechanisms, while IMM-TSF includes a specialized fusion module that supports asynchronous integration and realistic evaluation. Experimental results demonstrate that explicit modeling of irregular multimodal time series data can significantly improve forecasting performance.

Takeaways, Limitations

•
Takeaways:
◦
Bridging the gap between research and practical application by presenting a new dataset and benchmark library that reflects the characteristics of real-world time series data.
◦
Demonstrates the potential to improve forecasting performance through a specialized fusion module for irregular, multi-modal time series data.
◦
Expand the scope of research by systematically classifying various types of irregularities and promote time series analysis research in realistic environments.
•
Limitations:
◦
Only results for a specific dataset (Time-IMM) are presented, so further verification of generalization performance on other datasets is needed.
◦
There may be a lack of consideration for the complexity and computational cost of the provided fusion module.
◦
Further research is needed to validate performance in specific use cases and real-world applications.
👍
Made with Slashpage