Breaking Silos: Adaptive Model Fusion Unlocks Better Time Series Forecasting

Time-series forecasting plays a critical role in many real-world applications. Although increasingly powerful models have been developed and achieved superior results on benchmark datasets, through a fine-grained sample-level inspection, we find that (i) no single model consistently outperforms others across different test samples, but instead (ii) each model excels in specific cases. These findings prompt us to explore how to adaptively leverage the distinct strengths of various forecasting models for different samples. We introduce TimeFuse, a framework for collective time-series forecasting with sample-level adaptive fusion of heterogeneous models. TimeFuse utilizes meta-features to characterize input time series and trains a learnable fusor to predict optimal model fusion weights for any given input. The fusor can leverage samples from diverse datasets for joint training, allowing it to adapt to a wide variety of temporal patterns and thus generalize to new inputs, even from unseen datasets. Extensive experiments demonstrate the effectiveness of TimeFuse in various long-/short-term forecasting tasks, achieving near-universal improvement over the state-of-the-art individual models. Code is available at https://github.com/ZhiningLiu1998/TimeFuse.

Forecasting future trends (like energy demand, traffic patterns, or weather changes) is crucial in many real-world settings. While many powerful forecasting models have been developed, none consistently work best across all scenarios. Different models perform better on different types of data. Our work introduces a new approach called TIMEFUSE that smartly combines the strengths of multiple forecasting models for each individual case, rather than relying on a one-size-fits-all solution. TIMEFUSE learns to recognize the unique patterns in each input and selects the best combination of models accordingly. This adaptive strategy improves forecasting accuracy and works well even on new, unseen data. Our method is efficient, interpretable, and outperforms existing techniques across a wide range of forecasting tasks.