Deep Non-Parametric Time Series Forecaster

TL;DR

This paper introduces non-parametric time series forecasting models that generate predictions by sampling from empirical distributions, ensuring stable and reasonable forecasts without assuming specific parametric forms.

Contribution

It proposes a novel non-parametric forecasting approach that does not rely on parametric assumptions and includes a global version that learns sampling strategies across multiple series.

Findings

Models produce stable, reasonable forecasts across datasets.

Global version effectively learns sampling strategies from related series.

Performance is consistent, establishing strong baseline methods.

Abstract

This paper presents non-parametric baseline models for time series forecasting. Unlike classical forecasting models, the proposed approach does not assume any parametric form for the predictive distribution and instead generates predictions by sampling from the empirical distribution according to a tunable strategy. By virtue of this, the model is always able to produce reasonable forecasts (i.e., predictions within the observed data range) without fail unlike classical models that suffer from numerical stability on some data distributions. Moreover, we develop a global version of the proposed method that automatically learns the sampling strategy by exploiting the information across multiple related time series. The empirical evaluation shows that the proposed methods have reasonable and consistent performance across all datasets, proving them to be strong baselines to be considered in one's forecasting toolbox.