how to handle non-stationary time series

[JasonAlexTan]

Great work on TimeMixer and TimeMixer++! I've been using them for electricity market forecasting (predicting electricity futures prices), and they've been performing really well. One issue I’m dealing with is the heavy non-stationarity in electricity data. Do you have any advice or best practices for handling non-stationary time series with your models?

[kevinliu2000]

Thank you for being so interested in this work! For handling non-stationary time series, we recommend checking out “Non-stationary Transformers: Exploring the Stationarity in Time Series Forecasting.” Furthermore, you’re welcome to submit a merge request with your implementation of processing electricity data—we’d be happy to review it and learn from your approach.

[JasonAlexTan]

Thanks for the recommendation! I put together a pull request with my implementation. Appreciate your openness to feedback and collaboration.