lbfgs | Limited-Memory BFGS | Base64 library

You need to implement a dedicated get_feature_names function, as you are using a custom transformer.

Please refer to this question for details, where you can find a code example.

cross_val_score is meant for scoring a model by cross-validation, if you do:

Both iris_features and iris_labels are arrays and array literals in OCaml are delimited with the [|, |] style parentheses, e.g.,

We should import make_pipeline from imblearn.pipeline and not from sklearn.pipeline: make_pipeline from sklearn needs the transformers to implement fit and transform methods. sklearn.pipeline import Pipeline was conflicting with imblearn.pipeline import Pipeline!

You're using LogisticRegression, which is a special case of Linear Regression used for categorical dependent variables.

This is not necessarily wrong, as you might intend to do so, but that means you need sufficient data per category and enough iterations for the model to converge (which your error points out, it hasn't done).

I suspect, however, that what you intended to use is LinearRegression (used for continuous dependent variables) from sklearn library.

You can use the warnings-module to temporarily suppress warnings. Either all warnings or specific warnings.

In this case scikit-learn is raising a ConvergenceWarning so I suggest suppressing exactly that type of warning. That warning-class is located in sklearn.exceptions.ConvergenceWarning so import it beforehand and use the context-manager catch_warnings and the function simplefilter to ignore the warning, i.e. not print it to the screen:

As you are using regression, the model will attempt to predict a continuous value, which is how you are training it (note that the output Y is a single value):

• from X = [1,2,3] fit Y = 0
• from X = [2,1,3] fit Y = 1
• from X = [2,1,2] fit Y = 2

The output you are expecting is the one for a classifier, where each class gets a probability as output, i.e. confidence of the prediction. You should use a classification model instead, if that's what you want/need. And training it accordingly (each index in the output represents a class)

• from X = [1,2,3] fit Y = [1, 0, 0]
• from X = [2,1,3] fit Y = [0, 1, 0]
• from X = [2,1,2] fit Y = [0, 0, 1]

This is the gradient. Take the partials wrt each of the variables.

We can numerically check the gradient is correct.

Recalled that I can achieve this via Shap Values using something along the following (but using LinearExplainer instead of TreeExplainer):