EMTSF:Extraordinary Mixture of SOTA Models for Time Series Forecasting

TL;DR

This paper introduces EMTSF, a mixture of state-of-the-art models integrated via a Transformer-based MoE framework, achieving superior time series forecasting performance on standard benchmarks.

Contribution

It presents a novel MoE framework combining multiple SOTA TSF models within a Transformer architecture, outperforming existing methods.

Findings

EMTSF surpasses all existing TSF models on benchmarks.

The mixture of diverse models enhances forecasting accuracy.

The approach effectively leverages recent insights into TSF data characteristics.

Abstract

The immense success of the Transformer architecture in Natural Language Processing has led to its adoption in Time Se ries Forecasting (TSF), where superior performance has been shown. However, a recent important paper questioned their effectiveness by demonstrating that a simple single layer linear model outperforms Transformer-based models. This was soon shown to be not as valid, by a better transformer-based model termed PatchTST. More re cently, TimeLLM demonstrated even better results by repurposing a Large Language Model (LLM) for the TSF domain. Again, a follow up paper challenged this by demonstrating that removing the LLM component or replacing it with a basic attention layer in fact yields better performance. One of the challenges in forecasting is the fact that TSF data favors the more recent past, and is sometimes subject to unpredictable events. Based upon these recent insights in TSF, we propose a strong Mixture of Experts (MoE) framework. Our method combines the state-of-the-art (SOTA) models including xLSTM, en hanced Linear, PatchTST, and minGRU, among others. This set of complimentary and diverse models for TSF are integrated in a Trans former based MoE gating network. Our proposed model outperforms all existing TSF models on standard benchmarks, surpassing even the latest approaches based on MoE frameworks.