Support for AdaBoostClassifier (revisited)

[MooTong123]

I am going to generate pmml file for adaboost model, but it failed. I found a similar problem in 2017. (jpmml/jpmml-sklearn#41) Does the latest version of sklearn2pmml support adaboost now?

[vruusmann]

Does the latest version of sklearn2pmml support adaboost now?

The list of supported SkLearn transformer and model types is given in JPMML-SkLearn's README file: https://github.com/jpmml/jpmml-sklearn#supported-packages

The sklearn.ensemble.AdaBoostRegressor class is listed, but the AdaBoostClassifier class isn't.

I found a similar problem in 2017 (jpmml/jpmml-sklearn#41)

I checked my comments there, and they appear to be relevant today as well.

In short, it should be possible to support AdaBoostClassifier.predict(X), but not AdaBoostClassifier.predict_proba(X).

Will your application scenario be satisfied with AdaBoostClassifier.predict(X) alone?

The first thing to do would be to create a small test script that verifies the earlier claim that predict(X) and predict_proba(X) methods make conflicting predictions. If it's still the case, then I'd say that this model type does not deserve any further attention. There are better (ie. "more logical") boosted classifiers available.