CODEWITHSUNDEEP
pythonh2o
ceharep
ceharep

Reputation: 439

H2O - getting cross-validation results from python code

I'm trying to access the results of some H2O models using python.

I specifically want the cross-validation results. I'm able to get r2 and mae using the code below. I'd ideally like the standard deviation scores too.

I can see the data using .cross_validation_metrics_summary , but can't work out how to return the specific values (e.g. cross validation sd)

import h2o 

h2o.init()

def get_model_det(current_model):
    r2_score = current_model.r2(xval = "TRUE")
    mae_score = current_model.mae(xval = "True")
    varimp = current_model.varimp()
    print(current_model.cross_validation_metrics_summary)
    print(r2_score, mae_score)

current_model = h2o.get_model("XGBoost_2_AutoML_20200513_153924")
get_model_det(current_model)

Upvotes: 1

Views: 218

Answers (1)

Neema Mashayekhi
Neema Mashayekhi

Reputation: 930

If you would like to call out the specific values from cross_validation_metrics_summary, you can use the following:

current_model.cross_validation_metrics_summary().as_data_frame()[['', 'sd']]

The last part [['', 'sd']] will call the two columns of interest. '' is the name of each score (e.g. accuracy, auc) and 'sd' would give their corresponding standard deviations.

Outputs a table:

+-------+----------+--------------+
| index |    ''    |      sd      |
+-------+----------+--------------+
| 0     | accuracy | 0.0048520584 |
| 1     | auc      | 0.011593064  |
| 2     | aucpr    | 0.011920754  |
| ...   | ...      | ...          |
+-------+----------+--------------+

Upvotes: 1

Related Questions