Authors: Gerald Woo, Chenghao Liu, Akshat Kumar, Caiming Xiong, Silvio Savarese, Doyen Sahoo

Published on: February 04, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2402.02592

Summary

• What is new: Introduction of the Moirai architecture and LOTSA, achieving universal forecasting with a single model.
• Why this is important: Time series forecasting’s limitation by a one-model-per-dataset framework.
• What the research proposes: Moirai, a novel Transformer architecture enhanced for universal forecasting, trained on the LOTSA.
• Results: Moirai demonstrates competitive or superior zero-shot forecasting performance.

Technical Details

Technological frameworks used: Transformer architecture

Models used: Masked Encoder-based Universal Time Series Forecasting Transformer (Moirai)

Data used: Large-scale Open Time Series Archive (LOTSA)

Potential Impact

Potential impacts across finance, supply chain, weather forecasting, and more.

Want to implement this idea in a business?

We have generated a startup concept here: ForeCastAI.