H2O AutoML -如何提供权重

Question

下面是我的例子，Default日期集来自ISLR包。数据不平衡，所以我重新平衡它，只运行H2O AutoML与gbm。

library(ISLR)
library(h2o)
library(magrittr)
library(dplyr)
core_count <- detectCores()
h2o.init(nthreads = (core_count -1))
my_df <- Default
x <- setdiff(colnames(df_train), 'default')
y <- 'default'

my_df %<>% mutate(weights = if_else(default =='No',
0.6/table(my_df$default)[[1]],0.4/table(my_df$default)[[2]]))
aml_test <- h2o.automl(x = x, y = y,
training_frame = as.h2o(my_df[1:8000, ]),
validation_frame = as.h2o(my_df[8001:10000, ]),
nfolds = 0, 
weights_column = "weights",
include_algos = c('GBM'),
seed = 12345,
max_runtime_secs = 1200)

生成以下错误:

09:46:49.611: Skipping training of model GBM_1_AutoML_20210821_094649 due to exception:
water.exceptions.H2OModelBuilderIllegalArgumentException: Illegal argument(s) for GBM model:
GBM_1_AutoML_20210821_094649.  Details: ERRR on field: _min_rows: The dataset size is too 
small to split for min_rows=1.0: must have at least 2.0 (weighted) rows, but have only 
0.7172904568994339.

09:46:49.622: Skipping training of model GBM_2_AutoML_20210821_094649 due to exception:
water.exceptions.H2OModelBuilderIllegalArgumentException: Illegal argument(s) for GBM model: 
GBM_2_AutoML_20210821_094649.  Details: ERRR on field: _min_rows: The dataset size is too 
small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 
0.7172904568994339.

09:46:49.630: Skipping training of model GBM_3_AutoML_20210821_094649 due to exception: 
water.exceptions.H2OModelBuilderIllegalArgumentException: Illegal argument(s) for GBM model: 
GBM_3_AutoML_20210821_094649.  Details: ERRR on field: _min_rows: The dataset size is too 
small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 
0.7172904568994339.

09:46:49.637: Skipping training of model GBM_4_AutoML_20210821_094649 due to exception: 
water.exceptions.H2OModelBuilderIllegalArgumentException: Illegal argument(s) for GBM model: 
GBM_4_AutoML_20210821_094649.  Details: ERRR on field: _min_rows: The dataset size is too 
small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 
0.7172904568994339.

09:46:49.644: Skipping training of model GBM_5_AutoML_20210821_094649 due to exception: 
water.exceptions.H2OModelBuilderIllegalArgumentException: Illegal argument(s) for GBM model: 
GBM_5_AutoML_20210821_094649.  Details: ERRR on field: _min_rows: The dataset size is too 
small to split for min_rows=100.0: must have at least 200.0 (weighted) rows, but have only 
0.7172904568994339.
|===================================================================================| 100%


09:49:50.241: Empty leaderboard.
AutoML was not able to build any model within a max runtime constraint of 1200 seconds,
you may want to increase this value before retrying.The leaderboard contains zero models: 
try running AutoML for longer (the default is 1 hour).

本质上，无论何时提供类的权重，它都不能与GBM一起工作。没有重量也能很好地工作。它甚至没有运行整整20分钟。没有模型生成。

Answer 1

您的输出中出现错误消息

Details: ERRR on field: _min_rows: The dataset size is too 
small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 
0.7xxxx.

看起来您需要增加权重值和/或增加行数。试着将你的体重列乘以10或100倍，看看是否有帮助。我怀疑这不会是一个问题，如果你尝试设置权重列为所有的。