{AALF}: Almost Always Linear Forecasting

Recent work for time-series forecasting increasingly leverages the high predictive power of Deep Learning models. With this increase in model complexity, however, comes a lack in understanding of the underlying model decision process, which is problematic for safety-critical application scenarios. At the same time, simple, interpretable forecasting methods such as {ARIMA} and {ETS} still perform very well, sometimes on-par with Deep Learning approaches. We argue that using interpretable forecasters leads to good predictions in most cases. However, the forecasting performance can be improved by selecting a Deep Learning method only for few, important predictions, increasing the overall interpretability of the forecasting process. In this context, we propose a novel online model selection framework which learns to identify these predictions. An extensive empirical study on various real-world datasets containing over 3500 individual time-series shows that our selection methodology performs comparable to state-of-the-art online model selection methods in most cases while being significantly more interpretable. We find that almost always choosing a simple autoregressive or exponential smoothing model for forecasting, results in competitive performance, suggesting that the need for opaque black-box models in time-series forecasting might be smaller than recent works would suggest.