Authors: Peng Chen, Yingying Zhang, Yunyao Cheng, Yang Shu, Yihang Wang, Qingsong Wen, Bin Yang, Chenjuan Guo

Published on: February 04, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2402.05956

Summary

• What is new: Introduction of multi-scale transformers with adaptive pathways for time series forecasting, capturing varying scale characteristics.
• Why this is important: Existing methods struggle with modeling time series at different scales, limiting understanding of various characteristics.
• What the research proposes: The proposed Pathformer integrates multi-scale division, dual attention, and adaptive pathways to model time series effectively at various scales.
• Results: Pathformer surpasses all current models in performance on eleven real-world datasets and shows stronger generalization in different scenarios.

Technical Details

Technological frameworks used: Transformer-based models

Models used: Pathformer

Data used: Eleven real-world datasets

Potential Impact

Finance, supply chain, weather forecasting, energy management, and companies reliant on time series forecasting technologies.

Want to implement this idea in a business?

We have generated a startup concept here: ForecastFlow.