DDG-DA: Data Distribution Generation for Predictable Concept Drift Adaptation

In many real-world scenarios, we often deal with streaming data that is sequentially collected over time. Due to the non-stationary nature of environment, distribution may change in unpredictable ways, which known as concept drift literature. To handle drift, previous methods first detect when/where happens and then adapt models fit latest data. However, there are still cases some underlying factors environment evolution predictable, making it possible model future trend data, while such not fully explored work. this paper, propose a novel method DDG-DA, can effectively forecast improve performance models. Specifically, train predictor estimate distribution, leverage generate training samples, finally on generated We conduct experiments three tasks (forecasting stock price trend, electricity load solar irradiance) obtained significant improvement multiple widely-used