About mimic explainer: about interpret-community

[gaugup]

Hi @GISH123,

Sorry for the late response. If you are using the LightGBM model, then hopefully you will have "feature_importances_" that can give you ordering of the features in your trained lightgbm model.

However, the explainers under mimic explainer also give you local feature importance per sample of your evaluation data. So you could train lightgbm model and explain it using the mimic explainer explainer with lightgbm as the surrogate model to get both local and global feature importance. This is a perfectly valid scenario. However, the surrogate model might be lossy but for the scenario that you described the explanations provided by lightgbm surrogate model for your lightgbm model should be very close.

You can read more about the blackbox explainers and surrogate models on the link https://christophm.github.io/interpretable-ml-book/global.html.

Let us know if you have more questions.

Regards,

[gaugup]

@GISH123, I hope my above response helped with your understanding of blackbox explainers. Since there is no response on this issue for sometime, I am closing this issue. IF you have further questions, please feel free to open another issue.

Regards