UrbanAI 2025 Challenge: Linear vs Transformer Models for Long-Horizon Exogenous Temperature Forecasting
This work addresses temperature forecasting for urban AI applications, but it is incremental as it compares existing methods in a specific setting.
The study tackled long-horizon exogenous-only temperature forecasting using linear and Transformer models, finding that linear baselines, especially DLinear, consistently outperformed more complex Transformers across all splits.
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.