Training and Evaluating Causal Forecasting Models for Time-Series
This addresses the need for reliable causal forecasting in decision-making applications, representing an incremental advance in adapting existing frameworks to time-series.
The paper tackles the problem of time-series models failing to generalize to out-of-distribution forecasting scenarios, particularly for causal effects of actions, by extending orthogonal statistical learning to train causal models and using Regression Discontinuity Designs for evaluation, resulting in improved generalization.
Deep learning time-series models are often used to make forecasts that inform downstream decisions. Since these decisions can differ from those in the training set, there is an implicit requirement that time-series models will generalize outside of their training distribution. Despite this core requirement, time-series models are typically trained and evaluated on in-distribution predictive tasks. We extend the orthogonal statistical learning framework to train causal time-series models that generalize better when forecasting the effect of actions outside of their training distribution. To evaluate these models, we leverage Regression Discontinuity Designs popular in economics to construct a test set of causal treatment effects.