Model Selection

Options

Hello, I'm getting this error when trying to do Model Selection for backwards regression on a GLM: java.lang.ArrayIndexOutOfBoundsException. Does anyone know what this means?

Answers

  • wendycwong
    wendycwong Member Posts: 1
    Options

    I remember fixing this error before. Please try again using the latest H2O-3 version, 3.40.0.4. Which version of H2O-3 did you see this error in? If you can provide code, I will try to reproduce the error. Thanks.

  • stuartking
    stuartking Member Posts: 6
    Options

    @wendycwong I'm running into the same error. Difficult to provide fully reproducible code, but here is the model orchestration and call. Does anything stand out for why this might be happening? I'm using H2O-3 version 3.46.0.1.

    from h2o.estimators import H2OModelSelectionEstimator
    
    model = H2OModelSelectionEstimator(
        max_predictor_number=6,
        mode="backward",
        family="tweedie",
        solver='IRLSM',
        tweedie_variance_power=1.4,
        tweedie_link_power=0,
        lambda_=0,
        remove_collinear_columns=True,
        seed=12345
    )
    
    model.train(
        x=predictors, 
        y=target, 
        training_frame=X_train,
        weights_column=weight
    )