Do Stationarity Transformations Actually Improve Time Series Forecasts? A Controlled Experimental Evaluation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Malla, Bhanu Suraj, Hu, Yuqing
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916021262090240
author Malla, Bhanu Suraj
Hu, Yuqing
author_facet Malla, Bhanu Suraj
Hu, Yuqing
contents Stationarity transformations are standard preprocessing in time series forecasting, yet their actual impact on accuracy across different non-stationarity types and model families has received little controlled evaluation. We construct synthetic datasets with known properties - trend, seasonality, heteroscedasticity, and combinations - and apply fourteen transformation configurations across seven models and three forecast horizons (3,528 experiments). Stationarity is quantified via consensus ratios from ten statistical tests, and each transform-dataset pair is classified as matched or mismatched based on whether the transform targets the dataset's known non-stationarity. For matched pairs, transforms improve forecasts only 18% of the time. The primary exception is variance stabilization: log and Box-Cox on heteroscedastic data improve accuracy in 60-65% of cases. Differencing a linear-trend series - a textbook use case - worsens forecasts in all cases tested. Mediation analysis confirms that while transforms achieve trend stationarity, this does not translate into lower forecast error; the mechanism is signal attenuation. Real-world validation on TSA airport passenger data corroborates these findings. Our results suggest transformation selection should be guided by empirical out-of-sample evaluation rather than theoretical stationarity assumptions.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17689
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Do Stationarity Transformations Actually Improve Time Series Forecasts? A Controlled Experimental Evaluation
Malla, Bhanu Suraj
Hu, Yuqing
Methodology
Statistics Theory
62M10, 62P30
G.3; D.2.13; I.6.4
Stationarity transformations are standard preprocessing in time series forecasting, yet their actual impact on accuracy across different non-stationarity types and model families has received little controlled evaluation. We construct synthetic datasets with known properties - trend, seasonality, heteroscedasticity, and combinations - and apply fourteen transformation configurations across seven models and three forecast horizons (3,528 experiments). Stationarity is quantified via consensus ratios from ten statistical tests, and each transform-dataset pair is classified as matched or mismatched based on whether the transform targets the dataset's known non-stationarity. For matched pairs, transforms improve forecasts only 18% of the time. The primary exception is variance stabilization: log and Box-Cox on heteroscedastic data improve accuracy in 60-65% of cases. Differencing a linear-trend series - a textbook use case - worsens forecasts in all cases tested. Mediation analysis confirms that while transforms achieve trend stationarity, this does not translate into lower forecast error; the mechanism is signal attenuation. Real-world validation on TSA airport passenger data corroborates these findings. Our results suggest transformation selection should be guided by empirical out-of-sample evaluation rather than theoretical stationarity assumptions.
title Do Stationarity Transformations Actually Improve Time Series Forecasts? A Controlled Experimental Evaluation
topic Methodology
Statistics Theory
62M10, 62P30
G.3; D.2.13; I.6.4
url https://arxiv.org/abs/2605.17689