I am new to time-series modeling, and I was wondering what the standard way of quantifying feature importances are in a time-series setting? What types of models allow for the greatest interpretation of the feature space?

I am looking for something, which does not necessarily function like the Random Forest Regressor's feature importance call, but provides a similar insight.


For time series data,

  1. Sensitivity analysis can help with overall Importance of a feature. For example, is "Day of the week" a good feature for stock price forecasting. LIME is one approach that can help. Details : https://arxiv.org/abs/1606.05386 . One simple way is to mask each feature and check the impact on model's performance.
  2. Auto-corelation and Seasonality removal (Details in tutorial at end of the answer)
  3. SHAP : (SHapley Additive exPlanations) is good at identifying features that impact output with lag (https://medium.com/datadriveninvestor/time-step-wise-feature-importance-in-deep-learning-using-shap-e1c46a655455)

End to end example : https://machinelearningmastery.com/feature-selection-time-series-forecasting-python/


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.