SHAP¶
-
ExpyBox.
shap
()¶ Create dialog for shap, providing force and decision plots
- Returns
None
Method parameters¶
- Plot
Which plot to draw decision, force or both
- Background data
What background data will be used to sample from, when simulating “missing” feature
KMeans: use KMeans to sample from provided dataset
Custom variable: provide variable (that is in
kernel_globals
) with instances to use
- Count of KMeans centers
Number of means to use when creating background data. Only used if
Background data
is set toKMeans
- Background data variable
Variable with background data from which the “missing” features will be sampled. Only used if
Background data
is set toCustom variable
- Link
A generalized linear model link to connect the feature importance values to the model output
Since the feature importance values, phi, sum up to the model output, it often makes sense to connect them to the ouput with a link function where link(outout) = sum(phi).
If the model output is a probability then the LogitLink link function makes the feature importance values have log-odds units.
- Model sample size
Number of times to re-evaluate the model when explaining each prediction.
More samples lead to lower variance estimates of the SHAP values.
- Auto choose model sample size
The
auto
setting usesModel sample size
= 2 * train_data.shape[1] + 2048.
- L1 regularization
The l1 regularization to use for feature selection (the estimation procedure is based on a debiased lasso).
You can choose following inputs:
The
auto
option currently uses “aic” when less that 20% of the possible sample space is enumerated, otherwise it uses no regularization.The
aic
orbic
options use the AIC and BIC rules for regularizationInteger
selects a fix number of top featuresfloat
directly sets the “alpha” parameter of the sklearn.linear_model.Lasso model used for feature selection
- Class to plot
If explaining classification problem, select which class prediction to explain