some helpful ideas for H2O

H2O-usage.py

import h2o
import pandas
Inputdaten = pandas.Dataframe()
PREDICT_COL_NAME = 'tbd'


def pddf2h2odf(pddf: pandas.DataFrame) -> h2o.H2OFrame:
    # %% Sometimes, H2O adds duplicated to pandas-dataframes
    # Solution from https://stackoverflow.com/a/59588925/617339
    srcindexnameforlaterdeletion = pddf.index.name
    h2odf = h2o.H2OFrame(pddf.reset_index())
    if len(h2odf) != len(pddf):
        reimport = h2odf.as_data_frame()
        h2odf = h2odf.drop(reimport[reimport.duplicated()].index, axis=0)
    return h2odf.drop(srcindexnameforlaterdeletion)


with h2o.backend.H2OLocalServer.start(verbose=__debug__, name="H2O-Cluster") as hs:
    if __debug__:
        print("==========SERVER-INFO==========")
        print(hs)
        print("===============================")
    with h2o.connect(server=hs, verbose=__debug__) as hc:
        if __debug__:
            print("==========CONNECTION-INFO==========")
            print(hc)
            print("===================================")
        DATA = pddf2h2odf(Inputdaten)
        # http://archive.is/hBLPg
        print(DATA.describe())
        #FEATURES = DATA.columns.remove(PREDICT_COL_NAME)  # : List[Any]
        # features.remove(PREDICT_COL_NAME)
        aml = h2o.automl.H2OAutoML(seed=1)  # max_models=MAX_MODELS,
        aml.train(y=PREDICT_COL_NAME, training_frame=DATA)#x=FEATURES, 
        print(aml.leader)
        Inputdaten['H2O'] = aml.predict(test_data=DATA).as_data_frame().values
print(Inputdaten[[PREDICT_COL_NAME, 'H2O']])