默认参数

m <-
  h2o.gbm(
    x = x,
    y = y,
    training_frame = train,
    model_id = "GBM",
    nfolds = 10,
    validation_frame = valid
  )

h2o.varimp(m)

h2o.performance(m, test)

调参数

m1 <-
  h2o.grid("gbm",
           grid_id = "GBM_grid",
           search_criteria = list(strategy = "RandomDiscrete",max_model=50),
           hyper_params = list(
             max_depth = c(5,20,50),
             min_rows = c(2,5,10),
             sample_rate = c(0.5,0.8,0.95,1),
             cod_sample_rate = c(0.5,0.8,0.95,1),
             cod_sample_rate_per_tree = c(0.8,0.99,1),
             learn_rate = c(0,1)
             seed=1
           ),
           x=x,y=y,training_frame=train,validation_frame=valid,
           stopping_tolerance = 0.001,
           stopping_rounds = 3,
           score_tree_interval = 10,
           ntrees = 400)