Goal is to automate building and tuning a classification model to predict employee attrition, using the h2o::h2o.automl.
library(h2o)## Warning: package 'h2o' was built under R version 4.2.3##
## ----------------------------------------------------------------------
##
## Your next step is to start H2O:
## > h2o.init()
##
## For H2O package documentation, ask for help:
## > ??h2o
##
## After starting H2O, you can use the Web UI at http://localhost:54321
## For more information visit https://docs.h2o.ai
##
## ----------------------------------------------------------------------##
## Attaching package: 'h2o'## The following objects are masked from 'package:stats':
##
## cor, sd, var## The following objects are masked from 'package:base':
##
## &&, %*%, %in%, ||, apply, as.factor, as.numeric, colnames,
## colnames<-, ifelse, is.character, is.factor, is.numeric, log,
## log10, log1p, log2, round, signif, trunclibrary(tidyverse)## Warning: package 'readr' was built under R version 4.2.3## Warning: package 'dplyr' was built under R version 4.2.3## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.0
## ✔ ggplot2 3.4.4 ✔ tibble 3.2.1
## ✔ lubridate 1.9.2 ✔ tidyr 1.3.0
## ✔ purrr 1.0.2## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ lubridate::day() masks h2o::day()
## ✖ dplyr::filter() masks stats::filter()
## ✖ lubridate::hour() masks h2o::hour()
## ✖ dplyr::lag() masks stats::lag()
## ✖ lubridate::month() masks h2o::month()
## ✖ lubridate::week() masks h2o::week()
## ✖ lubridate::year() masks h2o::year()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorslibrary(tidymodels)## ── Attaching packages ────────────────────────────────────── tidymodels 1.1.1 ──
## ✔ broom 1.0.5 ✔ rsample 1.2.0
## ✔ dials 1.2.0 ✔ tune 1.1.2
## ✔ infer 1.0.6 ✔ workflows 1.1.3
## ✔ modeldata 1.3.0 ✔ workflowsets 1.0.1
## ✔ parsnip 1.1.1 ✔ yardstick 1.3.0
## ✔ recipes 1.0.9## Warning: package 'infer' was built under R version 4.2.3## Warning: package 'modeldata' was built under R version 4.2.3## Warning: package 'recipes' was built under R version 4.2.3## Warning: package 'yardstick' was built under R version 4.2.3## ── Conflicts ───────────────────────────────────────── tidymodels_conflicts() ──
## ✖ scales::discard() masks purrr::discard()
## ✖ dplyr::filter() masks stats::filter()
## ✖ recipes::fixed() masks stringr::fixed()
## ✖ dplyr::lag() masks stats::lag()
## ✖ yardstick::spec() masks readr::spec()
## ✖ recipes::step() masks stats::step()
## • Use tidymodels_prefer() to resolve common conflicts.library(tidyquant)## Loading required package: PerformanceAnalytics
## Loading required package: xts
## Loading required package: zoo
##
## Attaching package: 'zoo'
##
## The following objects are masked from 'package:base':
##
## as.Date, as.Date.numeric
##
##
## Attaching package: 'xts'
##
## The following objects are masked from 'package:dplyr':
##
## first, last
##
##
## Attaching package: 'PerformanceAnalytics'
##
## The following object is masked from 'package:graphics':
##
## legend
##
## Loading required package: quantmod
## Loading required package: TTR
##
## Attaching package: 'TTR'
##
## The following object is masked from 'package:dials':
##
## momentum
##
## Registered S3 method overwritten by 'quantmod':
## method from
## as.zoo.data.frame zooexpedition <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-09-22/members.csv')## Rows: 76519 Columns: 21
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (10): expedition_id, member_id, peak_id, peak_name, season, sex, citizen...
## dbl (5): year, age, highpoint_metres, death_height_metres, injury_height_me...
## lgl (6): hired, success, solo, oxygen_used, died, injured
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.factors_vec <- expedition %>% select(year, age, highpoint_metres, death_height_metres, injury_height_metres) %>% names()
data_clean <- expedition %>%
# Drop Variables
select(-c(death_height_metres, injury_height_metres, death_cause, injury_type)) %>%
# Drop Observations with missing values
drop_na() %>%
# Mutate Logical Variables
mutate(died = case_when(died == "TRUE" ~ "died", died == "FALSE" ~ "no")) %>%
mutate(across(where(is.logical), factor))set.seed(1234)
data_split <- initial_split(expedition, strata = "died")
train_tbl <- training(data_split)
test_tbl <- testing(data_split)recipe_obj <- recipe(died ~ ., data = train_tbl) %>%
# Remove zero variance variables
step_zv(all_predictors()) # Initialize H2o
h2o.init()## Connection successful!
##
## R is connected to the H2O cluster:
## H2O cluster uptime: 6 days 23 hours
## H2O cluster timezone: America/New_York
## H2O data parsing timezone: UTC
## H2O cluster version: 3.44.0.3
## H2O cluster version age: 4 months and 9 days
## H2O cluster name: H2O_started_from_R_Reed_fyb567
## H2O cluster total nodes: 1
## H2O cluster total memory: 0.04 GB
## H2O cluster total cores: 8
## H2O cluster allowed cores: 8
## H2O cluster healthy: TRUE
## H2O Connection ip: localhost
## H2O Connection port: 54321
## H2O Connection proxy: NA
## H2O Internal Security: FALSE
## R Version: R version 4.2.2 (2022-10-31)## Warning in h2o.clusterInfo():
## Your H2O cluster version is (4 months and 9 days) old. There may be a newer version available.
## Please download and install the latest version from: https://h2o-release.s3.amazonaws.com/h2o/latest_stable.htmlsplit.h2o <- h2o.splitFrame(as.h2o(train_tbl), ratios = c(0.85), seed = 2567)##
|
| | 0%
|
|======================================================================| 100%train_h2o <- split.h2o[[1]]
valid_h2o <- split.h2o[[2]]
test_h2o <- as.h2o(test_tbl)##
|
| | 0%
|
|======================================================================| 100%y <- "died"
x <- setdiff(names(train_tbl), y)
auto_ml_models_h2o <- h2o.automl(
x = x,
y = y,
training_frame = train_h2o,
validation_frame = valid_h2o,
leaderboard_frame = test_h2o,
max_runtime_secs = 30,
nfolds = 5,
seed = 3456
)##
|
| | 0%
|
|=== | 4%
## 10:48:49.611: User specified a validation frame with cross-validation still enabled. Please note that the models will still be validated using cross-validation only, the validation frame will be used to provide purely informative validation metrics on the trained models.
## 10:48:49.664: _train param, Dropping bad and constant columns: [member_id, peak_name, death_cause, peak_id, sex, citizenship, expedition_role, season, expedition_id, injury_type]
|
|======== | 12%
|
|============== | 19%
## 10:48:56.534: _train param, Dropping bad and constant columns: [member_id, peak_name, death_cause, peak_id, sex, citizenship, expedition_role, season, expedition_id, injury_type]
|
|=================== | 27%
|
|=========================== | 38%
|
|================================ | 46%
|
|====================================== | 55%
## 10:49:04.959: _train param, Dropping bad and constant columns: [member_id, peak_name, death_cause, peak_id, sex, citizenship, expedition_role, season, expedition_id, injury_type]
|
|============================================= | 64%
## 10:49:07.691: GBM_1_AutoML_23_20240430_104849 [GBM def_5] failed: water.exceptions.H2OModelBuilderIllegalArgumentException: Illegal argument(s) for GBM model: GBM_1_AutoML_23_20240430_104849_cv_1. Details: ERRR on field: _ntrees: The tree model will not fit in the driver node's memory (1.0 KB per tree x 10000 > Zero ) - try decreasing ntrees and/or max_depth or increasing min_rows!
##
## 10:49:07.716: _train param, Dropping unused columns: [member_id, peak_name, death_cause, peak_id, sex, citizenship, expedition_role, season, expedition_id, injury_type]
|
|================================================== | 71%
|
|======================================================= | 79%
|
|============================================================== | 88%
|
|=================================================================== | 95%
## 10:49:18.403: _train param, Dropping bad and constant columns: [member_id, peak_name, death_cause, peak_id, sex, citizenship, expedition_role, season, expedition_id, injury_type]
|
|======================================================================| 100%auto_ml_models_h2o@leaderboard## model_id auc
## 1 StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849 0.9926976
## 2 XGBoost_1_AutoML_23_20240430_104849 0.9926976
## 3 GLM_1_AutoML_23_20240430_104849 0.6598395
## logloss aucpr mean_per_class_error rmse mse
## 1 0.003824179 0.98755542 0.01666667 0.02167596 0.0004698474
## 2 0.006979407 0.98755542 0.01666667 0.02218753 0.0004922865
## 3 0.072959431 0.04539119 0.41118966 0.11779778 0.0138763175
##
## [3 rows x 7 columns]auto_ml_models_h2o@leader## Model Details:
## ==============
##
## H2OBinomialModel: stackedensemble
## Model ID: StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849
## Model Summary for Stacked Ensemble:
## key value
## 1 Stacking strategy cross_validation
## 2 Number of base models (used / total) 1/2
## 3 # XGBoost base models (used / total) 1/1
## 4 # GLM base models (used / total) 0/1
## 5 Metalearner algorithm GLM
## 6 Metalearner fold assignment scheme Random
## 7 Metalearner nfolds 5
## 8 Metalearner fold_column NA
## 9 Custom metalearner hyperparameters None
##
##
## H2OBinomialMetrics: stackedensemble
## ** Reported on training data. **
##
## MSE: 0.0006041881
## RMSE: 0.02458024
## LogLoss: 0.004874646
## Mean Per-Class Error: 0.01986755
## AUC: 0.9920701
## AUCPR: 0.9780595
## Gini: 0.9841403
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 9769 0 0.000000 =0/9769
## TRUE 6 145 0.039735 =6/151
## Totals 9775 145 0.000605 =6/9920
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.979730 14
## 2 max f2 0.990079 0.967957 14
## 3 max f0point5 0.990079 0.991792 14
## 4 max accuracy 0.990079 0.999395 14
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 52
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.979630 14
## 9 max min_per_class_accuracy 0.001192 0.980132 19
## 10 max mean_per_class_accuracy 0.001192 0.986125 19
## 11 max tns 0.999268 9769.000000 0
## 12 max fns 0.999268 136.000000 0
## 13 max fps 0.000523 9769.000000 53
## 14 max tps 0.000530 151.000000 52
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.900662 0
## 17 max fpr 0.000523 1.000000 53
## 18 max tpr 0.000530 1.000000 52
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: stackedensemble
## ** Reported on validation data. **
##
## MSE: 0.0004607463
## RMSE: 0.021465
## LogLoss: 0.003949776
## Mean Per-Class Error: 0.0141844
## AUC: 0.9922614
## AUCPR: 0.9790211
## Gini: 0.9845228
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 8536 0 0.000000 =0/8536
## TRUE 4 137 0.028369 =4/141
## Totals 8540 137 0.000461 =4/8677
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.985612 12
## 2 max f2 0.990079 0.977175 12
## 3 max f0point5 0.990079 0.994194 12
## 4 max accuracy 0.990079 0.999539 12
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 49
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.985483 12
## 9 max min_per_class_accuracy 0.001427 0.978723 15
## 10 max mean_per_class_accuracy 0.001427 0.987546 15
## 11 max tns 0.999268 8536.000000 0
## 12 max fns 0.999268 122.000000 0
## 13 max fps 0.000523 8536.000000 50
## 14 max tps 0.000530 141.000000 49
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.865248 0
## 17 max fpr 0.000523 1.000000 50
## 18 max tpr 0.000530 1.000000 49
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: stackedensemble
## ** Reported on cross-validation data. **
## ** 5-fold cross-validation on training data (Metrics computed for combined holdout predictions) **
##
## MSE: 0.0005154047
## RMSE: 0.02270253
## LogLoss: 0.00451254
## Mean Per-Class Error: 0.01798561
## AUC: 0.9830846
## AUCPR: 0.9712374
## Gini: 0.9661691
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 48017 0 0.000000 =0/48017
## TRUE 25 670 0.035971 =25/695
## Totals 48042 670 0.000513 =25/48712
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.961868 0.981685 216
## 2 max f2 0.961868 0.971014 216
## 3 max f0point5 0.961868 0.992593 216
## 4 max accuracy 0.961868 0.999487 216
## 5 max precision 0.997918 1.000000 0
## 6 max recall 0.000471 1.000000 399
## 7 max specificity 0.997918 1.000000 0
## 8 max absolute_mcc 0.961868 0.981594 216
## 9 max min_per_class_accuracy 0.001278 0.971223 256
## 10 max mean_per_class_accuracy 0.001662 0.982996 238
## 11 max tns 0.997918 48017.000000 0
## 12 max fns 0.997918 694.000000 0
## 13 max fps 0.000471 48017.000000 399
## 14 max tps 0.000471 695.000000 399
## 15 max tnr 0.997918 1.000000 0
## 16 max fnr 0.997918 0.998561 0
## 17 max fpr 0.000471 1.000000 399
## 18 max tpr 0.000471 1.000000 399
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## Cross-Validation Metrics Summary:
## mean sd cv_1_valid cv_2_valid cv_3_valid cv_4_valid
## accuracy 0.999486 0.000282 0.999693 0.999380 0.999797 0.999483
## auc 0.985555 0.013253 0.991555 0.987596 0.999918 0.984481
## err 0.000514 0.000282 0.000307 0.000620 0.000203 0.000517
## err_count 5.000000 2.738613 3.000000 6.000000 2.000000 5.000000
## f0point5 0.992461 0.004409 0.995671 0.990415 0.997089 0.993103
## cv_5_valid
## accuracy 0.999077
## auc 0.964224
## err 0.000923
## err_count 9.000000
## f0point5 0.986025
##
## ---
## mean sd cv_1_valid cv_2_valid cv_3_valid
## precision 1.000000 0.000000 1.000000 1.000000 1.000000
## r2 0.963066 0.020856 0.978234 0.953183 0.985225
## recall 0.963689 0.020634 0.978723 0.953846 0.985611
## residual_deviance 87.436554 40.447500 60.203827 98.165150 43.186370
## rmse 0.022000 0.006274 0.017597 0.024906 0.014344
## specificity 1.000000 0.000000 1.000000 1.000000 1.000000
## cv_4_valid cv_5_valid
## precision 1.000000 1.000000
## r2 0.965844 0.932844
## recall 0.966443 0.933824
## residual_deviance 87.124510 148.502910
## rmse 0.022760 0.030392
## specificity 1.000000 1.000000best_model <- auto_ml_models_h2o@leader
best_model## Model Details:
## ==============
##
## H2OBinomialModel: stackedensemble
## Model ID: StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849
## Model Summary for Stacked Ensemble:
## key value
## 1 Stacking strategy cross_validation
## 2 Number of base models (used / total) 1/2
## 3 # XGBoost base models (used / total) 1/1
## 4 # GLM base models (used / total) 0/1
## 5 Metalearner algorithm GLM
## 6 Metalearner fold assignment scheme Random
## 7 Metalearner nfolds 5
## 8 Metalearner fold_column NA
## 9 Custom metalearner hyperparameters None
##
##
## H2OBinomialMetrics: stackedensemble
## ** Reported on training data. **
##
## MSE: 0.0006041881
## RMSE: 0.02458024
## LogLoss: 0.004874646
## Mean Per-Class Error: 0.01986755
## AUC: 0.9920701
## AUCPR: 0.9780595
## Gini: 0.9841403
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 9769 0 0.000000 =0/9769
## TRUE 6 145 0.039735 =6/151
## Totals 9775 145 0.000605 =6/9920
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.979730 14
## 2 max f2 0.990079 0.967957 14
## 3 max f0point5 0.990079 0.991792 14
## 4 max accuracy 0.990079 0.999395 14
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 52
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.979630 14
## 9 max min_per_class_accuracy 0.001192 0.980132 19
## 10 max mean_per_class_accuracy 0.001192 0.986125 19
## 11 max tns 0.999268 9769.000000 0
## 12 max fns 0.999268 136.000000 0
## 13 max fps 0.000523 9769.000000 53
## 14 max tps 0.000530 151.000000 52
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.900662 0
## 17 max fpr 0.000523 1.000000 53
## 18 max tpr 0.000530 1.000000 52
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: stackedensemble
## ** Reported on validation data. **
##
## MSE: 0.0004607463
## RMSE: 0.021465
## LogLoss: 0.003949776
## Mean Per-Class Error: 0.0141844
## AUC: 0.9922614
## AUCPR: 0.9790211
## Gini: 0.9845228
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 8536 0 0.000000 =0/8536
## TRUE 4 137 0.028369 =4/141
## Totals 8540 137 0.000461 =4/8677
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.985612 12
## 2 max f2 0.990079 0.977175 12
## 3 max f0point5 0.990079 0.994194 12
## 4 max accuracy 0.990079 0.999539 12
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 49
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.985483 12
## 9 max min_per_class_accuracy 0.001427 0.978723 15
## 10 max mean_per_class_accuracy 0.001427 0.987546 15
## 11 max tns 0.999268 8536.000000 0
## 12 max fns 0.999268 122.000000 0
## 13 max fps 0.000523 8536.000000 50
## 14 max tps 0.000530 141.000000 49
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.865248 0
## 17 max fpr 0.000523 1.000000 50
## 18 max tpr 0.000530 1.000000 49
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: stackedensemble
## ** Reported on cross-validation data. **
## ** 5-fold cross-validation on training data (Metrics computed for combined holdout predictions) **
##
## MSE: 0.0005154047
## RMSE: 0.02270253
## LogLoss: 0.00451254
## Mean Per-Class Error: 0.01798561
## AUC: 0.9830846
## AUCPR: 0.9712374
## Gini: 0.9661691
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 48017 0 0.000000 =0/48017
## TRUE 25 670 0.035971 =25/695
## Totals 48042 670 0.000513 =25/48712
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.961868 0.981685 216
## 2 max f2 0.961868 0.971014 216
## 3 max f0point5 0.961868 0.992593 216
## 4 max accuracy 0.961868 0.999487 216
## 5 max precision 0.997918 1.000000 0
## 6 max recall 0.000471 1.000000 399
## 7 max specificity 0.997918 1.000000 0
## 8 max absolute_mcc 0.961868 0.981594 216
## 9 max min_per_class_accuracy 0.001278 0.971223 256
## 10 max mean_per_class_accuracy 0.001662 0.982996 238
## 11 max tns 0.997918 48017.000000 0
## 12 max fns 0.997918 694.000000 0
## 13 max fps 0.000471 48017.000000 399
## 14 max tps 0.000471 695.000000 399
## 15 max tnr 0.997918 1.000000 0
## 16 max fnr 0.997918 0.998561 0
## 17 max fpr 0.000471 1.000000 399
## 18 max tpr 0.000471 1.000000 399
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## Cross-Validation Metrics Summary:
## mean sd cv_1_valid cv_2_valid cv_3_valid cv_4_valid
## accuracy 0.999486 0.000282 0.999693 0.999380 0.999797 0.999483
## auc 0.985555 0.013253 0.991555 0.987596 0.999918 0.984481
## err 0.000514 0.000282 0.000307 0.000620 0.000203 0.000517
## err_count 5.000000 2.738613 3.000000 6.000000 2.000000 5.000000
## f0point5 0.992461 0.004409 0.995671 0.990415 0.997089 0.993103
## cv_5_valid
## accuracy 0.999077
## auc 0.964224
## err 0.000923
## err_count 9.000000
## f0point5 0.986025
##
## ---
## mean sd cv_1_valid cv_2_valid cv_3_valid
## precision 1.000000 0.000000 1.000000 1.000000 1.000000
## r2 0.963066 0.020856 0.978234 0.953183 0.985225
## recall 0.963689 0.020634 0.978723 0.953846 0.985611
## residual_deviance 87.436554 40.447500 60.203827 98.165150 43.186370
## rmse 0.022000 0.006274 0.017597 0.024906 0.014344
## specificity 1.000000 0.000000 1.000000 1.000000 1.000000
## cv_4_valid cv_5_valid
## precision 1.000000 1.000000
## r2 0.965844 0.932844
## recall 0.966443 0.933824
## residual_deviance 87.124510 148.502910
## rmse 0.022760 0.030392
## specificity 1.000000 1.000000auto_ml_models_h2o %>% typeof()## [1] "S4"auto_ml_models_h2o %>% slotNames()## [1] "project_name" "leader" "leaderboard" "event_log"
## [5] "modeling_steps" "training_info"auto_ml_models_h2o@leaderboard## model_id auc
## 1 StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849 0.9926976
## 2 XGBoost_1_AutoML_23_20240430_104849 0.9926976
## 3 GLM_1_AutoML_23_20240430_104849 0.6598395
## logloss aucpr mean_per_class_error rmse mse
## 1 0.003824179 0.98755542 0.01666667 0.02167596 0.0004698474
## 2 0.006979407 0.98755542 0.01666667 0.02218753 0.0004922865
## 3 0.072959431 0.04539119 0.41118966 0.11779778 0.0138763175
##
## [3 rows x 7 columns]auto_ml_models_h2o@leader## Model Details:
## ==============
##
## H2OBinomialModel: stackedensemble
## Model ID: StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849
## Model Summary for Stacked Ensemble:
## key value
## 1 Stacking strategy cross_validation
## 2 Number of base models (used / total) 1/2
## 3 # XGBoost base models (used / total) 1/1
## 4 # GLM base models (used / total) 0/1
## 5 Metalearner algorithm GLM
## 6 Metalearner fold assignment scheme Random
## 7 Metalearner nfolds 5
## 8 Metalearner fold_column NA
## 9 Custom metalearner hyperparameters None
##
##
## H2OBinomialMetrics: stackedensemble
## ** Reported on training data. **
##
## MSE: 0.0006041881
## RMSE: 0.02458024
## LogLoss: 0.004874646
## Mean Per-Class Error: 0.01986755
## AUC: 0.9920701
## AUCPR: 0.9780595
## Gini: 0.9841403
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 9769 0 0.000000 =0/9769
## TRUE 6 145 0.039735 =6/151
## Totals 9775 145 0.000605 =6/9920
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.979730 14
## 2 max f2 0.990079 0.967957 14
## 3 max f0point5 0.990079 0.991792 14
## 4 max accuracy 0.990079 0.999395 14
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 52
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.979630 14
## 9 max min_per_class_accuracy 0.001192 0.980132 19
## 10 max mean_per_class_accuracy 0.001192 0.986125 19
## 11 max tns 0.999268 9769.000000 0
## 12 max fns 0.999268 136.000000 0
## 13 max fps 0.000523 9769.000000 53
## 14 max tps 0.000530 151.000000 52
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.900662 0
## 17 max fpr 0.000523 1.000000 53
## 18 max tpr 0.000530 1.000000 52
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: stackedensemble
## ** Reported on validation data. **
##
## MSE: 0.0004607463
## RMSE: 0.021465
## LogLoss: 0.003949776
## Mean Per-Class Error: 0.0141844
## AUC: 0.9922614
## AUCPR: 0.9790211
## Gini: 0.9845228
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 8536 0 0.000000 =0/8536
## TRUE 4 137 0.028369 =4/141
## Totals 8540 137 0.000461 =4/8677
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.985612 12
## 2 max f2 0.990079 0.977175 12
## 3 max f0point5 0.990079 0.994194 12
## 4 max accuracy 0.990079 0.999539 12
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 49
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.985483 12
## 9 max min_per_class_accuracy 0.001427 0.978723 15
## 10 max mean_per_class_accuracy 0.001427 0.987546 15
## 11 max tns 0.999268 8536.000000 0
## 12 max fns 0.999268 122.000000 0
## 13 max fps 0.000523 8536.000000 50
## 14 max tps 0.000530 141.000000 49
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.865248 0
## 17 max fpr 0.000523 1.000000 50
## 18 max tpr 0.000530 1.000000 49
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: stackedensemble
## ** Reported on cross-validation data. **
## ** 5-fold cross-validation on training data (Metrics computed for combined holdout predictions) **
##
## MSE: 0.0005154047
## RMSE: 0.02270253
## LogLoss: 0.00451254
## Mean Per-Class Error: 0.01798561
## AUC: 0.9830846
## AUCPR: 0.9712374
## Gini: 0.9661691
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## FALSE TRUE Error Rate
## FALSE 48017 0 0.000000 =0/48017
## TRUE 25 670 0.035971 =25/695
## Totals 48042 670 0.000513 =25/48712
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.961868 0.981685 216
## 2 max f2 0.961868 0.971014 216
## 3 max f0point5 0.961868 0.992593 216
## 4 max accuracy 0.961868 0.999487 216
## 5 max precision 0.997918 1.000000 0
## 6 max recall 0.000471 1.000000 399
## 7 max specificity 0.997918 1.000000 0
## 8 max absolute_mcc 0.961868 0.981594 216
## 9 max min_per_class_accuracy 0.001278 0.971223 256
## 10 max mean_per_class_accuracy 0.001662 0.982996 238
## 11 max tns 0.997918 48017.000000 0
## 12 max fns 0.997918 694.000000 0
## 13 max fps 0.000471 48017.000000 399
## 14 max tps 0.000471 695.000000 399
## 15 max tnr 0.997918 1.000000 0
## 16 max fnr 0.997918 0.998561 0
## 17 max fpr 0.000471 1.000000 399
## 18 max tpr 0.000471 1.000000 399
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## Cross-Validation Metrics Summary:
## mean sd cv_1_valid cv_2_valid cv_3_valid cv_4_valid
## accuracy 0.999486 0.000282 0.999693 0.999380 0.999797 0.999483
## auc 0.985555 0.013253 0.991555 0.987596 0.999918 0.984481
## err 0.000514 0.000282 0.000307 0.000620 0.000203 0.000517
## err_count 5.000000 2.738613 3.000000 6.000000 2.000000 5.000000
## f0point5 0.992461 0.004409 0.995671 0.990415 0.997089 0.993103
## cv_5_valid
## accuracy 0.999077
## auc 0.964224
## err 0.000923
## err_count 9.000000
## f0point5 0.986025
##
## ---
## mean sd cv_1_valid cv_2_valid cv_3_valid
## precision 1.000000 0.000000 1.000000 1.000000 1.000000
## r2 0.963066 0.020856 0.978234 0.953183 0.985225
## recall 0.963689 0.020634 0.978723 0.953846 0.985611
## residual_deviance 87.436554 40.447500 60.203827 98.165150 43.186370
## rmse 0.022000 0.006274 0.017597 0.024906 0.014344
## specificity 1.000000 0.000000 1.000000 1.000000 1.000000
## cv_4_valid cv_5_valid
## precision 1.000000 1.000000
## r2 0.965844 0.932844
## recall 0.966443 0.933824
## residual_deviance 87.124510 148.502910
## rmse 0.022760 0.030392
## specificity 1.000000 1.000000predictions <- h2o.predict(best_model, newdata = test_h2o)##
|
| | 0%
|
|======================================================================| 100%predictions_tbl <- predictions %>%
as_tibble()
predictions_tbl %>%
bind_cols(test_tbl)## # A tibble: 19,130 × 24
## predict FALSE. TRUE. expedition_id member_id peak_id peak_name year season
## <fct> <dbl> <dbl> <chr> <chr> <chr> <chr> <dbl> <chr>
## 1 FALSE 0.999 6.48e-4 AMAD78301 AMAD7830… AMAD Ama Dabl… 1978 Autumn
## 2 FALSE 0.999 6.48e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 3 FALSE 0.999 6.72e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 4 FALSE 0.999 6.72e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 5 FALSE 0.999 6.72e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 6 FALSE 0.999 6.48e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 7 FALSE 0.999 6.48e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 8 FALSE 0.999 6.48e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 9 FALSE 0.999 6.72e-4 AMAD79101 AMAD7910… AMAD Ama Dabl… 1979 Spring
## 10 FALSE 0.999 6.48e-4 AMAD79301 AMAD7930… AMAD Ama Dabl… 1979 Autumn
## # ℹ 19,120 more rows
## # ℹ 15 more variables: sex <chr>, age <dbl>, citizenship <chr>,
## # expedition_role <chr>, hired <lgl>, highpoint_metres <dbl>, success <lgl>,
## # solo <lgl>, oxygen_used <lgl>, died <lgl>, death_cause <chr>,
## # death_height_metres <dbl>, injured <lgl>, injury_type <chr>,
## # injury_height_metres <dbl>?h2o.performance
performance_h2o <- h2o.performance(best_model, newdata = test_h2o)
typeof(performance_h2o)## [1] "S4"slotNames(performance_h2o)## [1] "algorithm" "on_train" "on_valid" "on_xval" "metrics"performance_h2o@metrics## $model
## $model$`__meta`
## $model$`__meta`$schema_version
## [1] 3
##
## $model$`__meta`$schema_name
## [1] "ModelKeyV3"
##
## $model$`__meta`$schema_type
## [1] "Key<Model>"
##
##
## $model$name
## [1] "StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849"
##
## $model$type
## [1] "Key<Model>"
##
## $model$URL
## [1] "/3/Models/StackedEnsemble_BestOfFamily_1_AutoML_23_20240430_104849"
##
##
## $model_checksum
## [1] "-995355060137004016"
##
## $frame
## $frame$name
## [1] "test_tbl_sid_8238_3"
##
##
## $frame_checksum
## [1] "678340420273909232"
##
## $description
## NULL
##
## $scoring_time
## [1] 1.714489e+12
##
## $predictions
## NULL
##
## $MSE
## [1] 0.0004698474
##
## $RMSE
## [1] 0.02167596
##
## $nobs
## [1] 19130
##
## $custom_metric_name
## NULL
##
## $custom_metric_value
## [1] 0
##
## $r2
## [1] 0.9662339
##
## $logloss
## [1] 0.003824179
##
## $AUC
## [1] 0.9926976
##
## $pr_auc
## [1] 0.9875554
##
## $Gini
## [1] 0.9853951
##
## $mean_per_class_error
## [1] 0.01666667
##
## $domain
## [1] "FALSE" "TRUE"
##
## $cm
## $cm$`__meta`
## $cm$`__meta`$schema_version
## [1] 3
##
## $cm$`__meta`$schema_name
## [1] "ConfusionMatrixV3"
##
## $cm$`__meta`$schema_type
## [1] "ConfusionMatrix"
##
##
## $cm$table
## Confusion Matrix: Row labels: Actual class; Column labels: Predicted class
## FALSE TRUE Error Rate
## FALSE 18860 0 0.0000 = 0 / 18,860
## TRUE 9 261 0.0333 = 9 / 270
## Totals 18869 261 0.0005 = 9 / 19,130
##
##
## $thresholds_and_metric_scores
## Metrics for Thresholds: Binomial metrics as a function of classification thresholds
## threshold f1 f2 f0point5 accuracy precision recall specificity
## 1 0.999268 0.205980 0.139514 0.393401 0.987507 1.000000 0.114815 1.000000
## 2 0.999240 0.235294 0.161290 0.434783 0.987768 1.000000 0.133333 1.000000
## 3 0.998555 0.383234 0.279720 0.608365 0.989232 1.000000 0.237037 1.000000
## 4 0.998499 0.402367 0.296167 0.627306 0.989441 1.000000 0.251852 1.000000
## 5 0.997793 0.872651 0.810706 0.944846 0.996811 1.000000 0.774074 1.000000
## absolute_mcc min_per_class_accuracy mean_per_class_accuracy tns fns fps tps
## 1 0.336717 0.114815 0.557407 18860 239 0 31
## 2 0.362904 0.133333 0.566667 18860 234 0 36
## 3 0.484227 0.237037 0.618519 18860 206 0 64
## 4 0.499182 0.251852 0.625926 18860 202 0 68
## 5 0.878395 0.774074 0.887037 18860 61 0 209
## tnr fnr fpr tpr idx
## 1 1.000000 0.885185 0.000000 0.114815 0
## 2 1.000000 0.866667 0.000000 0.133333 1
## 3 1.000000 0.762963 0.000000 0.237037 2
## 4 1.000000 0.748148 0.000000 0.251852 3
## 5 1.000000 0.225926 0.000000 0.774074 4
##
## ---
## threshold f1 f2 f0point5 accuracy precision recall specificity
## 56 0.000549 0.183442 0.358775 0.123223 0.875745 0.101098 0.988889 0.874125
## 57 0.000549 0.085797 0.189792 0.055426 0.702561 0.044844 0.988889 0.698462
## 58 0.000542 0.074095 0.166521 0.047648 0.651176 0.038489 0.988889 0.646341
## 59 0.000530 0.073422 0.165161 0.047203 0.647726 0.038127 0.988889 0.642842
## 60 0.000530 0.029189 0.069912 0.018445 0.061160 0.014811 1.000000 0.047720
## 61 0.000523 0.027835 0.066799 0.017580 0.014114 0.014114 1.000000 0.000000
## absolute_mcc min_per_class_accuracy mean_per_class_accuracy tns fns fps
## 56 0.295114 0.874125 0.931507 16486 3 2374
## 57 0.175120 0.698462 0.843676 13173 3 5687
## 58 0.155863 0.646341 0.817615 12190 3 6670
## 59 0.154691 0.642842 0.815865 12124 3 6736
## 60 0.026585 0.047720 0.523860 900 0 17960
## 61 0.000000 0.000000 0.500000 0 0 18860
## tps tnr fnr fpr tpr idx
## 56 267 0.874125 0.011111 0.125875 0.988889 55
## 57 267 0.698462 0.011111 0.301538 0.988889 56
## 58 267 0.646341 0.011111 0.353659 0.988889 57
## 59 267 0.642842 0.011111 0.357158 0.988889 58
## 60 270 0.047720 0.000000 0.952280 1.000000 59
## 61 270 0.000000 0.000000 1.000000 1.000000 60
##
## $max_criteria_and_metric_scores
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.990079 0.983051 17
## 2 max f2 0.001531 0.976419 19
## 3 max f0point5 0.990079 0.993151 17
## 4 max accuracy 0.990079 0.999530 17
## 5 max precision 0.999268 1.000000 0
## 6 max recall 0.000530 1.000000 59
## 7 max specificity 0.999268 1.000000 0
## 8 max absolute_mcc 0.990079 0.982958 17
## 9 max min_per_class_accuracy 0.001192 0.988889 23
## 10 max mean_per_class_accuracy 0.001427 0.991691 20
## 11 max tns 0.999268 18860.000000 0
## 12 max fns 0.999268 239.000000 0
## 13 max fps 0.000523 18860.000000 60
## 14 max tps 0.000530 270.000000 59
## 15 max tnr 0.999268 1.000000 0
## 16 max fnr 0.999268 0.885185 0
## 17 max fpr 0.000523 1.000000 60
## 18 max tpr 0.000530 1.000000 59
##
## $gains_lift_table
## Gains/Lift Table: Avg response rate: 1.41 %, avg score: 1.42 %
## group cumulative_data_fraction lower_threshold lift cumulative_lift
## 1 1 0.01092525 0.997793 70.851852 70.851852
## 2 2 0.02117094 0.001192 20.966364 46.709739
## 3 3 0.03193936 0.000991 0.000000 30.961448
## 4 4 0.04511239 0.000960 0.000000 21.920561
## 5 5 0.05101934 0.000790 0.000000 19.382628
## 6 6 0.12409827 0.000648 0.000000 7.968595
## 7 7 0.31123889 0.000549 0.000000 3.177266
## 8 8 0.95295348 0.000530 0.017315 1.049369
## 9 9 1.00000000 0.000523 0.000000 1.000000
## response_rate score cumulative_response_rate cumulative_score capture_rate
## 1 1.000000 0.998162 1.000000 0.998162 0.774074
## 2 0.295918 0.264726 0.659259 0.643215 0.214815
## 3 0.000000 0.001061 0.436989 0.426711 0.000000
## 4 0.000000 0.000961 0.309386 0.302390 0.000000
## 5 0.000000 0.000814 0.273566 0.267474 0.000000
## 6 0.000000 0.000670 0.112468 0.110358 0.000000
## 7 0.000000 0.000554 0.044844 0.044335 0.000000
## 8 0.000244 0.000531 0.014811 0.014837 0.011111
## 9 0.000000 0.000523 0.014114 0.014164 0.000000
## cumulative_capture_rate gain cumulative_gain kolmogorov_smirnov
## 1 0.774074 6985.185185 6985.185185 0.774074
## 2 0.988889 1996.636432 4570.973937 0.981572
## 3 0.988889 -100.000000 2996.144754 0.970649
## 4 0.988889 -100.000000 2092.056135 0.957288
## 5 0.988889 -100.000000 1838.262750 0.951296
## 6 0.988889 -100.000000 696.859496 0.877171
## 7 0.988889 -100.000000 217.726645 0.687351
## 8 1.000000 -98.268528 4.936917 0.047720
## 9 1.000000 -100.000000 0.000000 0.000000
##
## $residual_deviance
## [1] 146.3131
##
## $null_deviance
## [1] 2836.923
##
## $AIC
## [1] 150.3131
##
## $loglikelihood
## [1] 0
##
## $null_degrees_of_freedom
## [1] 19129
##
## $residual_degrees_of_freedom
## [1] 19128h2o.auc(performance_h2o)## [1] 0.9926976h2o.confusionMatrix(performance_h2o)## Confusion Matrix (vertical: actual; across: predicted) for max f1 @ threshold = 0.990079190848319:
## FALSE TRUE Error Rate
## FALSE 18860 0 0.000000 =0/18860
## TRUE 9 261 0.033333 =9/270
## Totals 18869 261 0.000470 =9/19130h2o.metric(performance_h2o) %>% as_tibble() %>% filter(threshold %>% between(0.41, 0.42))## # A tibble: 0 × 20
## # ℹ 20 variables: threshold <dbl>, f1 <dbl>, f2 <dbl>, f0point5 <dbl>,
## # accuracy <dbl>, precision <dbl>, recall <dbl>, specificity <dbl>,
## # absolute_mcc <dbl>, min_per_class_accuracy <dbl>,
## # mean_per_class_accuracy <dbl>, tns <dbl>, fns <dbl>, fps <dbl>, tps <dbl>,
## # tnr <dbl>, fnr <dbl>, fpr <dbl>, tpr <dbl>, idx <int>