H2O for python not converging

By Kyoungmun Chang posted 10-19-2020 07:36

​Hello, everyone

I use both Rapidminer and Python H2O.
Rapidminer has H2O Deep Learning operator whose version is

When I do binary classification using H2O Deep Learning Operator of Rapidminer (using simply default parameters of H2O Deep Learning) 
its performance is perfect
Precision = Sensitivity = 100%, AUC = 100%

But when I do the same binary classification using Python H2O (H2ODeepLearningEstimator)
with the same data for Rapidminer,
it is not properly trained,
Precision is about 30%, Sensitivity is about 60~70%
I want Precision to be more than 50%, but I failed.

I modified and tuned arguments in H2ODeepLearningEstimator
such as weight initialization, l1, l2, max_w2, adaptive learning rate, acceleration function, bernoulli distribution, hidden layers, epochs etc
But I failed to train the dataset.

Could anyone of you help me with this difficult situation?

Have a nice day all of you​