h2O model and its configurations

Hi! I would like to build a predictive model for classification. It should be a binary classification that aims to predict which patients are “DIFICULT” or “EASY” to have a specific disease.

Is the H2O model in some way different than the normal machine learning algorithms? I’m implementing the Random Forest nodes (learner and predictor) to do a prediction about my patients. Also, with H2O model I have the option to use the Random Forest algorithm, which I have also implemented. I have compared both models and they give different results. In the H2O binomial scorer, I can specify which is the target class I would like to predict. In my case, my interest is to have a good prediction of the patients classified as “Difícil”.
image
However, in the scorer node I can not specify the target class, just the column I would like to predict.
image
Is this fact giving different results in the prediction?

I have another doubt regarding the previous question. In my workflow, I have optimised the parameters for each model and then I have compared all the models. For example, from the picture it can be seen that I have built a H2O Random Forest. With the node “Parameter Optimisation Loop End”, I have registered the parameters from the model that give the best accuracy. Then I have written also the name of the model thorugh the node “Constant Value Column”
image

The output from the Constant Value Column node is the following:

As I said, my interest is to predict appropiately the patients classified as “Difícil”. In the ouput table, I would like to add information about the difficult patients well predicted for the best accuracy model.

Blockquote

Is there any possibility I can do that?

Thank you in advance.
Helena Fortuny

To clarify, I would like to know the sensibility, the accuracy and the specificity for every iteration on each model. So at the end of Parameter Optimisation Loop End, I would like to have the parameters of the model (as I already have), the sensibility, the accuracy and the specificity.

Have you taken into account that partitions are different between h2o and “normal” model?
br

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.