1

I have a MOJO model that I want to explore for model metrics (rmse,roc, etc.) I understand all model metrics are available for a binary model, but I want to get these metrics from a MOJO model.

Input - Mojo Model and training dataset Output - Extract the model metrics

Tried extracting the metrics from a binary model using : ModelMetricsSupport.modelMetrics [ ModelMetricsSupervised] (binaryModel,valid)

But unable to do the same for a MOJO Model.

curios
  • 11
  • 2
  • please post a fully reproducible code snippet and list the version of H2O-3 and sparkling water you are using. Please note that the ability to see model metrics in addition to scoring was first introduced in H2O-3 version 3.26 https://www.h2o.ai/blog/h2o-release-3-26-yau/ – Lauren Aug 30 '19 at 22:48

1 Answers1

0

I was able to get model metrics from a MOJO model in Python(H2O), it can be done either ways : a. The exported zip of the mojo model has model json that contains all the metrics information. Path : /experimental/modelDetails.json

b. Second way is to import and invoke model_performance()

import h2o

h2o.init()

model = h2o.import_mojo("/Users/sonamsharma/delete/GLM_model_python_1567539104058_1.zip")

model.model_performance()

This can be either used as is or json can also be extracted.

For JAVA, i tried taking reference of the below issue fix and the code, but was unable to do that. https://github.com/h2oai/h2o-3/pull/3642

curios
  • 11
  • 2