Zero-shot Multivariate Time Series Forecasting Using Tabular Prior Fitted Networks

TL;DR

This paper introduces a novel zero-shot multivariate time series forecasting framework using tabular foundation models, recasting the problem as scalar regressions and demonstrating competitive performance with state-of-the-art methods.

Contribution

The paper presents a general framework for multivariate time series forecasting with tabular models, overcoming the limitation of treating channels independently.

Findings

Effective zero-shot forecasting with TabPFN-TS backbone.

Competitive performance against state-of-the-art tabular methods.

Recasting multivariate forecasting as scalar regression problems.

Abstract

Tabular foundation models, particularly Prior-data Fitted Networks like TabPFN have emerged as the leading contender in a myriad of tasks ranging from data imputation to label prediction on the tabular data format surpassing the historical successes of tree-based models. This has led to investigations on their applicability to forecasting time series data which can be formulated as a tabular problem. While recent work to this end has displayed positive results, most works have limited their treatment of multivariate time series problems to several independent univariate time series forecasting subproblems, thus ignoring any inter-channel interactions. Overcoming this limitation, we introduce a generally applicable framework for multivariate time series forecasting using tabular foundation models. We achieve this by recasting the multivariate time series forecasting problem as a series of scalar regression problems which can then be solved zero-shot by any tabular foundation model with regression capabilities. We present results of our method using the TabPFN-TS backbone and compare performance with the current state of the art tabular methods.