Q: Feature Contributions for Timeseries analysis

[lefig]

Hi there,

I use the fabulous time series analysis models and would appreciate a little help to understand what support there is (if any) to understand:

• Feature weights, Importance and Contributions (as with regression and classification learners)
• PermutationMetrics

Much obliged

Fig

[najeeb-kazmi]

Hi @lefig - thanks for reaching out to us. Currently, there is no support for feature contributions or permutation feature importance for time series. Both these features are supported for predictors, while time series algorithms are implemented as transforms.

Adding @gvashishtha to help triage this feature request.

[lefig]

I really appreciate this. Splendid news and thank you too.

Also I think that the following list of supported learners for feature contribution is up to date but perhaps someone may wish to review and check.

https://docs.microsoft.com/en-us/dotnet/api/microsoft.ml.transforms.featurecontributioncalculatingestimator?view=ml-dotnet