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.

Latent Factorization of Tensors with Threshold Distance Weighted Loss for Traffic Data Estimation

Created by
  • Haebom

Author

Lei Yang

Outline

In this paper, we propose a novel outlier-robust missing value correction model, TDWLFT, to address the missing problem of high-quality spatiotemporal traffic data, which is essential for improving the performance of intelligent transportation systems (ITS). Unlike the existing latent factor tensor decomposition (LFT) models that use the L2-norm, which is vulnerable to outliers, TDWLFT introduces the threshold distance weighting (TDW) loss function to reduce the influence of outliers. Experimental results show that TDWLFT outperforms existing state-of-the-art techniques in both prediction accuracy and computational efficiency on traffic speed datasets in various urban environments.

Takeaways, Limitations

•
Takeaways:
◦
Contributes to improving ITS performance by proposing a missing value correction technique for spatiotemporal traffic data that is robust to outliers.
◦
The TDWLFT model experimentally demonstrates higher prediction accuracy and computational efficiency than the conventional LFT model.
◦
The generalization performance of the model is verified through experiments in various urban environments.
•
Limitations:
◦
The performance of the proposed model may vary depending on the dataset used.
◦
Additional validation and optimization may be required for application to real ITS environments.
◦
It is possible that we only addressed robustness to certain types of outliers. Additional research may be needed for different types of outliers.
👍
Made with Slashpage