UrbanAI 2025 Challenge: Linear vs Transformer Models for Long-Horizon Exogenous Temperature Forecasting

TL;DR

This paper compares linear and Transformer models for long-horizon indoor temperature forecasting using only past temperature data, finding linear models outperform complex Transformers.

Contribution

It demonstrates that simple linear models can outperform Transformer architectures in long-horizon exogenous temperature forecasting tasks.

Findings

Linear models outperform Transformer models in accuracy.

DLinear achieves the best overall performance.

Linear models are strong baselines for exogenous temperature forecasting.

Abstract

We study long-horizon exogenous-only temperature forecasting - a challenging univariate setting where only the past values of the indoor temperature are used for prediction - using linear and Transformer-family models. We evaluate Linear, NLinear, DLinear, Transformer, Informer, and Autoformer under standardized train, validation, and test splits. Results show that linear baselines (Linear, NLinear, DLinear) consistently outperform more complex Transformer-family architectures, with DLinear achieving the best overall accuracy across all splits. These findings highlight that carefully designed linear models remain strong baselines for time series forecasting in challenging exogenous-only settings.