renanxcortes

# SHAP values: http://docs.h2o.ai/h2o/latest-stable/h2o-r/docs/reference/predict_contributions.H2OModel.html

SHAP_values <- predict_contributions.H2OModel(aml_leader, df_frame_split[[2]])

# Wrangling inspired here: https://bradleyboehmke.github.io/HOML/iml.html

shap_df <- SHAP_values %>%
  as.data.frame() %>%
  select(-BiasTerm) %>%
  gather(feature, shap_value) %>%

renanxcortes

# Metric for binary classification (deviance is the default). Check documentation here http://docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html
automl_model <- h2o.automl(#x = x, 
                           y = 'y',
                           balance_classes = TRUE,
                           training_frame = df_frame_split[[1]],
                           nfolds = 4,
                           #validation_frame = df_frame_split[[2]], # read help(h2o.automl) !!!Optional. This argument is ignored unless the user sets nfolds = 0!!!
                           leaderboard_frame = df_frame_split[[2]],
                           max_runtime_secs = 60 * 2, # Two minutes
                           #exclude_algos = "StackedEnsemble", # Global Importance of Stacked models is tricky

renanxcortes

renanxcortes

renanxcortes

renanxcortes

renanxcortes

renanxcortes