DynaConF: Dynamic Forecasting of Non-Stationary Time Series
This addresses forecasting challenges for domains with non-stationary data, but it is incremental as it builds on existing deep learning and Bayesian methods.
The paper tackles the problem of forecasting non-stationary time series by decoupling stationary conditional distribution modeling from non-stationary dynamics, resulting in a model that adapts better than state-of-the-art deep learning solutions on synthetic and real-world datasets.
Deep learning has shown impressive results in a variety of time series forecasting tasks, where modeling the conditional distribution of the future given the past is the essence. However, when this conditional distribution is non-stationary, it poses challenges for these models to learn consistently and to predict accurately. In this work, we propose a new method to model non-stationary conditional distributions over time by clearly decoupling stationary conditional distribution modeling from non-stationary dynamics modeling. Our method is based on a Bayesian dynamic model that can adapt to conditional distribution changes and a deep conditional distribution model that handles multivariate time series using a factorized output space. Our experimental results on synthetic and real-world datasets show that our model can adapt to non-stationary time series better than state-of-the-art deep learning solutions.