Import Data From module 7
data <- read_csv("../00_data/data_wrangled/data_clean2.csv") %>%
select(-notes) %>%
# h2o requires all variables to be either numeric or factors
mutate(across(where(is.character), factor))## New names:
## Rows: 7458 Columns: 9
## ── Column specification
## ──────────────────────────────────────────────────────── Delimiter: "," chr
## (4): coname, exec_fullname, ceo_dismissal, notes dbl (5): ...1,
## dismissal_dataset_id, tenure_no_ceodb, max_tenure_ceodb, fyea...
## 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.
## • `` -> `...1`set.seed(1234)
data_split <- initial_split(data, strata = "ceo_dismissal")
train_tbl <- training(data_split)
test_tbl <- testing(data_split)train_tbl <- train_tbl %>%
mutate(ceo_dismissal = as.factor(ceo_dismissal))
test_tbl <- test_tbl %>%
mutate(ceo_dismissal = as.factor(ceo_dismissal))recipe_obj <- recipe(ceo_dismissal ~ ., 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 minutes 19 seconds
## H2O cluster timezone: America/New_York
## H2O data parsing timezone: UTC
## H2O cluster version: 3.44.0.3
## H2O cluster version age: 1 year, 4 months and 15 days
## H2O cluster name: H2O_started_from_R_User_xzn825
## H2O cluster total nodes: 1
## H2O cluster total memory: 1.71 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 4.4.1 (2024-06-14 ucrt)## Warning in h2o.clusterInfo():
## Your H2O cluster version is (1 year, 4 months and 15 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.htmltrain_h2o <- as.h2o(train_tbl)## | | | 0% | |======================================================================| 100%test_h2o <- as.h2o(test_tbl)## | | | 0% | |======================================================================| 100%train_h2o[["ceo_dismissal"]] <- as.factor(train_h2o[["ceo_dismissal"]])
test_h2o[["ceo_dismissal"]] <- as.factor(test_h2o[["ceo_dismissal"]])
print(h2o.describe(train_h2o))## Label Type Missing Zeros PosInf NegInf Min Max
## 1 ...1 int 0 0 0 0 1 7475
## 2 dismissal_dataset_id int 0 0 0 0 2 559044
## 3 coname enum 0 1 0 0 0 3003
## 4 exec_fullname enum 0 1 0 0 0 5296
## 5 ceo_dismissal enum 0 1111 0 0 0 1
## 6 tenure_no_ceodb int 0 0 0 0 1 3
## 7 max_tenure_ceodb int 0 0 0 0 1 4
## 8 fyear_gone int 0 0 0 0 1980 2021
## Mean Sigma Cardinality
## 1 3747.5692830 2.144169e+03 NA
## 2 5660.7280529 2.682587e+04 NA
## 3 NA NA 3004
## 4 NA NA 5297
## 5 0.8013588 3.990129e-01 2
## 6 1.0262829 1.665610e-01 NA
## 7 1.0480958 2.300954e-01 NA
## 8 2006.4519936 7.514666e+00 NAprint(h2o.describe(test_h2o))## Label Type Missing Zeros PosInf NegInf Min Max
## 1 ...1 int 0 0 0 0 2 7461
## 2 dismissal_dataset_id int 0 0 0 0 1 559039
## 3 coname enum 0 1 0 0 0 1484
## 4 exec_fullname enum 0 1 0 0 0 1839
## 5 ceo_dismissal enum 0 371 0 0 0 1
## 6 tenure_no_ceodb int 0 0 0 0 1 3
## 7 max_tenure_ceodb int 0 0 0 0 1 3
## 8 fyear_gone int 0 0 0 0 1990 2021
## Mean Sigma Cardinality
## 1 3703.5324397 2.195641e+03 NA
## 2 5298.8632708 2.238609e+04 NA
## 3 NA NA 1485
## 4 NA NA 1840
## 5 0.8010724 3.993005e-01 2
## 6 1.0235925 1.553100e-01 NA
## 7 1.0466488 2.208809e-01 NA
## 8 2006.2482574 7.468390e+00 NAsplit.h2o <- h2o.splitFrame(data = train_h2o, ratio = c(0.85), seed = 2345)
train_h2o <- split.h2o[[1]]
valid_h2o <- split.h2o[[2]]
y <- "ceo_dismissal"
x <- setdiff(names(train_tbl), y)
models_h2o <-h2o.automl(
x = x,
y = y,
training_frame = train_h2o,
validation_frame = valid_h2o,
leaderboard_frame = test_h2o,
# max_runtime_secs = 30,
max_models = 10,
exclude_algos = "DeepLearning",
nfolds = 5,
seed = 3456
)## | | | 0% | |== | 3%
## 20:25:53.782: 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.
## 20:25:53.786: AutoML: XGBoost is not available; skipping it. | |=== | 4% | |======= | 10% | |========= | 12% | |=============== | 21% | |================== | 26% | |==================== | 29% | |======================= | 33% | |=============================================== | 67% | |======================================================================| 100%Examine the output of h2o.automl
models_h2o %>% typeof() ## [1] "S4"models_h2o %>% slotNames()## [1] "project_name" "leader" "leaderboard" "event_log"
## [5] "modeling_steps" "training_info"models_h2o@leaderboard## model_id auc logloss
## 1 GBM_1_AutoML_3_20250505_202553 0.5656065 0.5256068
## 2 StackedEnsemble_BestOfFamily_1_AutoML_3_20250505_202553 0.5584215 0.4949310
## 3 StackedEnsemble_AllModels_1_AutoML_3_20250505_202553 0.5430356 0.4991453
## 4 GBM_3_AutoML_3_20250505_202553 0.5420252 0.5085978
## 5 GBM_grid_1_AutoML_3_20250505_202553_model_1 0.5369934 0.5072509
## 6 GBM_2_AutoML_3_20250505_202553 0.5348348 0.5133798
## aucpr mean_per_class_error rmse mse
## 1 0.8254422 0.5000000 0.4055355 0.1644591
## 2 0.8373919 0.5000000 0.3976640 0.1581367
## 3 0.8286244 0.5000000 0.3995318 0.1596256
## 4 0.8189387 0.5000000 0.4025896 0.1620784
## 5 0.8104962 0.4973046 0.4019058 0.1615283
## 6 0.8126084 0.5000000 0.4049437 0.1639794
##
## [12 rows x 7 columns]models_h2o@leader## Model Details:
## ==============
##
## H2OBinomialModel: gbm
## Model ID: GBM_1_AutoML_3_20250505_202553
## Model Summary:
## number_of_trees number_of_internal_trees model_size_in_bytes min_depth
## 1 20 20 24694 8
## max_depth mean_depth min_leaves max_leaves mean_leaves
## 1 15 10.90000 28 32 29.65000
##
##
## H2OBinomialMetrics: gbm
## ** Reported on training data. **
##
## MSE: 0.0727113
## RMSE: 0.2696503
## LogLoss: 0.2641725
## Mean Per-Class Error: 0.06332186
## AUC: 0.9902714
## AUCPR: 0.9975053
## Gini: 0.9805428
## R^2: 0.5446184
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## dismissed not_dis Error Rate
## dismissed 847 100 0.105597 =100/947
## not_dis 80 3721 0.021047 =80/3801
## Totals 927 3821 0.037911 =180/4748
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.711267 0.976384 190
## 2 max f2 0.657988 0.983756 215
## 3 max f0point5 0.780842 0.981108 154
## 4 max accuracy 0.715440 0.962089 188
## 5 max precision 0.940063 1.000000 0
## 6 max recall 0.435188 1.000000 319
## 7 max specificity 0.940063 1.000000 0
## 8 max absolute_mcc 0.715440 0.881009 188
## 9 max min_per_class_accuracy 0.763147 0.950370 163
## 10 max mean_per_class_accuracy 0.780842 0.954696 154
## 11 max tns 0.940063 947.000000 0
## 12 max fns 0.940063 3798.000000 0
## 13 max fps 0.147692 947.000000 399
## 14 max tps 0.435188 3801.000000 319
## 15 max tnr 0.940063 1.000000 0
## 16 max fnr 0.940063 0.999211 0
## 17 max fpr 0.147692 1.000000 399
## 18 max tpr 0.435188 1.000000 319
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: gbm
## ** Reported on validation data. **
## ** Validation metrics **
##
## MSE: 0.1610166
## RMSE: 0.4012687
## LogLoss: 0.5169004
## Mean Per-Class Error: 0.4969512
## AUC: 0.5693743
## AUCPR: 0.8339982
## Gini: 0.1387486
## R^2: -0.0294209
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## dismissed not_dis Error Rate
## dismissed 1 163 0.993902 =163/164
## not_dis 0 681 0.000000 =0/681
## Totals 1 844 0.192899 =163/845
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.503794 0.893115 398
## 2 max f2 0.503794 0.954316 398
## 3 max f0point5 0.503794 0.839290 398
## 4 max accuracy 0.503794 0.807101 398
## 5 max precision 0.957102 1.000000 0
## 6 max recall 0.503794 1.000000 398
## 7 max specificity 0.957102 1.000000 0
## 8 max absolute_mcc 0.874719 0.124872 220
## 9 max min_per_class_accuracy 0.897750 0.548780 161
## 10 max mean_per_class_accuracy 0.878856 0.572445 214
## 11 max tns 0.957102 164.000000 0
## 12 max fns 0.957102 680.000000 0
## 13 max fps 0.495691 164.000000 399
## 14 max tps 0.503794 681.000000 398
## 15 max tnr 0.957102 1.000000 0
## 16 max fnr 0.957102 0.998532 0
## 17 max fpr 0.495691 1.000000 399
## 18 max tpr 0.503794 1.000000 398
##
## Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
## H2OBinomialMetrics: gbm
## ** Reported on cross-validation data. **
## ** 5-fold cross-validation on training data (Metrics computed for combined holdout predictions) **
##
## MSE: 0.1634281
## RMSE: 0.4042625
## LogLoss: 0.5201853
## Mean Per-Class Error: 0.4985476
## AUC: 0.5747655
## AUCPR: 0.8313301
## Gini: 0.149531
## R^2: -0.02352956
##
## Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
## dismissed not_dis Error Rate
## dismissed 3 944 0.996832 =944/947
## not_dis 1 3800 0.000263 =1/3801
## Totals 4 4744 0.199031 =945/4748
##
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.420150 0.889409 396
## 2 max f2 0.385576 0.952584 398
## 3 max f0point5 0.475385 0.834250 385
## 4 max accuracy 0.420150 0.800969 396
## 5 max precision 0.961763 1.000000 0
## 6 max recall 0.385576 1.000000 398
## 7 max specificity 0.961763 1.000000 0
## 8 max absolute_mcc 0.842146 0.118493 189
## 9 max min_per_class_accuracy 0.890740 0.546990 115
## 10 max mean_per_class_accuracy 0.866376 0.563691 156
## 11 max tns 0.961763 947.000000 0
## 12 max fns 0.961763 3799.000000 0
## 13 max fps 0.350050 947.000000 399
## 14 max tps 0.385576 3801.000000 398
## 15 max tnr 0.961763 1.000000 0
## 16 max fnr 0.961763 0.999474 0
## 17 max fpr 0.350050 1.000000 399
## 18 max tpr 0.385576 1.000000 398
##
## 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
## accuracy 0.801390 0.001197 0.803158 0.800000 0.801053
## auc 0.580013 0.019646 0.569574 0.575384 0.562185
## err 0.198610 0.001197 0.196842 0.200000 0.198947
## err_count 188.600000 1.140176 187.000000 190.000000 189.000000
## f0point5 0.834469 0.000957 0.835903 0.833333 0.834247
## f1 0.889618 0.000542 0.890323 0.888889 0.889538
## f2 0.952572 0.000221 0.952321 0.952381 0.952679
## lift_top_group 1.149263 0.137340 1.250000 1.250000 0.998686
## logloss 0.541275 0.006080 0.533277 0.542740 0.544452
## max_per_class_error 0.994731 0.009116 0.978947 1.000000 1.000000
## mcc 0.087119 0.031099 0.109109 NA NA
## mean_per_class_accuracy 0.502503 0.004274 0.509868 0.500000 0.500000
## mean_per_class_error 0.497497 0.004274 0.490132 0.500000 0.500000
## mse 0.166832 0.000786 0.166187 0.167723 0.167471
## pr_auc 0.837060 0.012549 0.840352 0.840541 0.818671
## precision 0.801352 0.001184 0.803175 0.800000 0.801053
## r2 -0.044853 0.005228 -0.038670 -0.048267 -0.050851
## recall 0.999737 0.000588 0.998684 1.000000 1.000000
## rmse 0.408450 0.000963 0.407661 0.409540 0.409233
## specificity 0.005269 0.009116 0.021053 0.000000 0.000000
## cv_4_valid cv_5_valid
## accuracy 0.801897 0.800843
## auc 0.579811 0.613113
## err 0.198103 0.199157
## err_count 188.000000 189.000000
## f0point5 0.834798 0.834065
## f1 0.889930 0.889409
## f2 0.952859 0.952620
## lift_top_group 0.998947 1.248684
## logloss 0.548699 0.537206
## max_per_class_error 0.994709 1.000000
## mcc 0.065129 NA
## mean_per_class_accuracy 0.502645 0.500000
## mean_per_class_error 0.497354 0.500000
## mse 0.166872 0.165908
## pr_auc 0.832832 0.852902
## precision 0.801688 0.800843
## r2 -0.046261 -0.040217
## recall 1.000000 1.000000
## rmse 0.408500 0.407318
## specificity 0.005291 0.000000?h2o.getModel## starting httpd help server ... done?h2o.saveModel
?h2o.loadModel
best_model <- models_h2o@leaderpredictions <- h2o.predict(best_model, newdata = test_h2o)## | | | 0% | |======================================================================| 100%## Warning in doTryCatch(return(expr), name, parentenv, handler): Test/Validation
## dataset column 'coname' has levels not trained on: ["ABIOMED INC", "ACUSON
## CORP", "ADESA INC", "ADTRAN INC", "ADVANCE CIRCUITS INC", "ADVANCED MARKETING
## SERVICES", "AEROFLEX INC", "AEROVIRONMENT INC", "ALASKA AIR GROUP INC", "ALBANY
## MOLECULAR RESH INC", ...403 not listed..., "WEATHERFORD ENTERRA INC", "WEBEX
## COMMUNICATIONS INC", "WELLMAN INC", "WESTERN GAS RESOURCES INC", "WESTROCK CO",
## "WEX INC", "WHEELABRATOR TECHNOLOGIES", "WYNDHAM DESTINATIONS INC", "WYNN
## RESORTS LTD", "YOUNKERS INC"]## Warning in doTryCatch(return(expr), name, parentenv, handler): Test/Validation
## dataset column 'exec_fullname' has levels not trained on: ["A. Frederick
## Gerstell", "A. George (Skip) Battle", "A. James Dearlove", "A. Lorne Weil", "A.
## William Reynolds", "Abe J. Gustin Jr.", "Adam D. Singer", "Adrian Adams",
## "Ahmad R. Chatila", "Alain C. Viry", ...1644 not listed..., "William Thomas
## Dillard", "William V. Hickey", "William W. Sprague Jr.", "William White Adams",
## "William Wrigley Jr.", "Willliam T. Jensen", "Wilson B. Sexton", "Woodson M.
## Hobbs", "Ying Lu", "Zan Guerry"]predictions_tbl <- predictions %>%
as_tibble()
predictions_tbl %>%
bind_cols(test_tbl)## New names:
## • `...1` -> `...4`## # A tibble: 1,865 × 11
## predict dismissed not_dis ...4 dismissal_dataset_id coname exec_fullname
## <fct> <dbl> <dbl> <dbl> <dbl> <fct> <fct>
## 1 not_dis 0.0992 0.901 2 12 AMERICAN … Robert L. Cr…
## 2 not_dis 0.103 0.897 3 13 AMERICAN … Donald J. Ca…
## 3 not_dis 0.241 0.759 9 65 AIR PRODU… Harold A. Wa…
## 4 not_dis 0.116 0.884 12 78 ALBERTSON… Gary Glenn M…
## 5 not_dis 0.127 0.873 13 80 ALCAN INC David Morton
## 6 not_dis 0.116 0.884 14 81 ALCAN INC Jacques Boug…
## 7 not_dis 0.129 0.871 17 88 ALEXANDER… Tinsley H. I…
## 8 not_dis 0.142 0.858 18 99 HONEYWELL… Lawrence A. …
## 9 not_dis 0.119 0.881 23 117 HESS CORP Leon Hess
## 10 not_dis 0.111 0.889 26 121 BEAM INC William J. A…
## # ℹ 1,855 more rows
## # ℹ 4 more variables: ceo_dismissal <fct>, tenure_no_ceodb <dbl>,
## # max_tenure_ceodb <dbl>, fyear_gone <dbl>?h2o.performance
performance_h2o <- h2o.performance(best_model, newdata = test_h2o)
confusion_matrix <- h2o.confusionMatrix(performance_h2o)
print(confusion_matrix)## Confusion Matrix (vertical: actual; across: predicted) for max f1 @ threshold = 0.229895324126849:
## dismissed not_dis Error Rate
## dismissed 0 371 1.000000 =371/371
## not_dis 0 1494 0.000000 =0/1494
## Totals 0 1865 0.198928 =371/1865#typeof(performance_h2o)
#slotNames(performance_h2o)
#performance_h2o@metrics
auc <- h2o.auc(performance_h2o)
print(paste("AUC:", auc))## [1] "AUC: 0.565606541169169"metrics <- performance_h2o@metrics
print(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] "GBM_1_AutoML_3_20250505_202553"
##
## $model$type
## [1] "Key<Model>"
##
## $model$URL
## [1] "/3/Models/GBM_1_AutoML_3_20250505_202553"
##
##
## $model_checksum
## [1] "447411804817538224"
##
## $frame
## $frame$name
## [1] "RTMP_sid_bc65_6"
##
##
## $frame_checksum
## [1] "9011093136619684768"
##
## $description
## NULL
##
## $scoring_time
## [1] 1.746491e+12
##
## $predictions
## NULL
##
## $MSE
## [1] 0.1644591
##
## $RMSE
## [1] 0.4055355
##
## $nobs
## [1] 1865
##
## $custom_metric_name
## NULL
##
## $custom_metric_value
## [1] 0
##
## $r2
## [1] -0.0320269
##
## $logloss
## [1] 0.5256068
##
## $AUC
## [1] 0.5656065
##
## $pr_auc
## [1] 0.8254422
##
## $Gini
## [1] 0.1312131
##
## $mean_per_class_error
## [1] 0.5
##
## $domain
## [1] "dismissed" "not_dis"
##
## $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
## dismissed not_dis Error Rate
## dismissed 0 371 1.0000 = 371 / 371
## not_dis 0 1494 0.0000 = 0 / 1,494
## Totals 0 1865 0.1989 = 371 / 1,865
##
##
## $thresholds_and_metric_scores
## Metrics for Thresholds: Binomial metrics as a function of classification thresholds
## threshold f1 f2 f0point5 accuracy precision recall specificity
## 1 0.953882 0.001337 0.000836 0.003329 0.198928 0.500000 0.000669 0.997305
## 2 0.952959 0.005337 0.003344 0.013210 0.200536 0.800000 0.002677 0.997305
## 3 0.951831 0.007995 0.005014 0.019711 0.201609 0.857143 0.004016 0.997305
## 4 0.950652 0.009315 0.005848 0.022876 0.201609 0.777778 0.004685 0.994609
## 5 0.949261 0.010631 0.006681 0.026008 0.201609 0.727273 0.005355 0.991914
## absolute_mcc min_per_class_accuracy mean_per_class_accuracy tns fns fps tps
## 1 0.024711 0.000669 0.498987 370 1493 1 1
## 2 0.000139 0.002677 0.499991 370 1490 1 4
## 3 0.008621 0.004016 0.500660 370 1488 1 6
## 4 0.004064 0.004685 0.499647 369 1487 2 7
## 5 0.014240 0.005355 0.498634 368 1486 3 8
## tnr fnr fpr tpr idx
## 1 0.997305 0.999331 0.002695 0.000669 0
## 2 0.997305 0.997323 0.002695 0.002677 1
## 3 0.997305 0.995984 0.002695 0.004016 2
## 4 0.994609 0.995315 0.005391 0.004685 3
## 5 0.991914 0.994645 0.008086 0.005355 4
##
## ---
## threshold f1 f2 f0point5 accuracy precision recall
## 395 0.447303 0.888425 0.950345 0.834080 0.799464 0.801399 0.996653
## 396 0.435700 0.889088 0.951378 0.834453 0.800536 0.801613 0.997992
## 397 0.389037 0.889154 0.951773 0.834265 0.800536 0.801289 0.998661
## 398 0.380633 0.889485 0.952290 0.834451 0.801072 0.801396 0.999331
## 399 0.362542 0.889220 0.952168 0.834078 0.800536 0.800966 0.999331
## 400 0.229895 0.889550 0.952685 0.834264 0.801072 0.801072 1.000000
## specificity absolute_mcc min_per_class_accuracy mean_per_class_accuracy tns
## 395 0.005391 0.013344 0.005391 0.501022 2
## 396 0.005391 0.026115 0.005391 0.501691 2
## 397 0.002695 0.013515 0.002695 0.500678 1
## 398 0.002695 0.024711 0.002695 0.501013 1
## 399 0.000000 0.011542 0.000000 0.499665 0
## 400 0.000000 0.000000 0.000000 0.500000 0
## fns fps tps tnr fnr fpr tpr idx
## 395 5 369 1489 0.005391 0.003347 0.994609 0.996653 394
## 396 3 369 1491 0.005391 0.002008 0.994609 0.997992 395
## 397 2 370 1492 0.002695 0.001339 0.997305 0.998661 396
## 398 1 370 1493 0.002695 0.000669 0.997305 0.999331 397
## 399 1 371 1493 0.000000 0.000669 1.000000 0.999331 398
## 400 0 371 1494 0.000000 0.000000 1.000000 1.000000 399
##
## $max_criteria_and_metric_scores
## Maximum Metrics: Maximum metrics at their respective thresholds
## metric threshold value idx
## 1 max f1 0.229895 0.889550 399
## 2 max f2 0.229895 0.952685 399
## 3 max f0point5 0.716851 0.835679 316
## 4 max accuracy 0.380633 0.801072 397
## 5 max precision 0.937262 0.880597 18
## 6 max recall 0.229895 1.000000 399
## 7 max specificity 0.953882 0.997305 0
## 8 max absolute_mcc 0.790605 0.107165 263
## 9 max min_per_class_accuracy 0.898800 0.555556 118
## 10 max mean_per_class_accuracy 0.899547 0.563519 116
## 11 max tns 0.953882 370.000000 0
## 12 max fns 0.953882 1493.000000 0
## 13 max fps 0.362542 371.000000 398
## 14 max tps 0.229895 1494.000000 399
## 15 max tnr 0.953882 0.997305 0
## 16 max fnr 0.953882 0.999331 0
## 17 max fpr 0.362542 1.000000 398
## 18 max tpr 0.229895 1.000000 399
##
## $gains_lift_table
## Gains/Lift Table: Avg response rate: 80.11 %, avg score: 87.30 %
## group cumulative_data_fraction lower_threshold lift cumulative_lift
## 1 1 0.01018767 0.944808 0.919820 0.919820
## 2 2 0.02037534 0.942024 1.182625 1.051222
## 3 3 0.03056300 0.939078 1.116924 1.073123
## 4 4 0.04021448 0.936127 1.109624 1.081883
## 5 5 0.05040214 0.934275 0.985521 1.062406
## 6 6 0.10026810 0.927785 0.953024 1.008007
## 7 7 0.15013405 0.923770 1.060407 1.025411
## 8 8 0.20000000 0.920008 1.100675 1.044177
## 9 9 0.30026810 0.913764 1.041385 1.043244
## 10 10 0.40053619 0.908046 1.034709 1.041108
## 11 11 0.50026810 0.900822 1.067118 1.046293
## 12 12 0.60000000 0.891748 0.953024 1.030790
## 13 13 0.69973190 0.879236 1.000004 1.026402
## 14 14 0.80000000 0.852719 0.981305 1.020750
## 15 15 0.89973190 0.776666 0.986581 1.016962
## 16 16 1.00000000 0.229895 0.847794 1.000000
## response_rate score cumulative_response_rate cumulative_score
## 1 0.736842 0.949343 0.736842 0.949343
## 2 0.947368 0.943407 0.842105 0.946375
## 3 0.894737 0.940356 0.859649 0.944369
## 4 0.888889 0.937267 0.866667 0.942664
## 5 0.789474 0.935080 0.851064 0.941131
## 6 0.763441 0.930628 0.807487 0.935908
## 7 0.849462 0.925622 0.821429 0.932492
## 8 0.881720 0.921697 0.836461 0.929800
## 9 0.834225 0.916837 0.835714 0.925472
## 10 0.828877 0.910654 0.834003 0.921762
## 11 0.854839 0.904460 0.838156 0.918313
## 12 0.763441 0.896665 0.825737 0.914714
## 13 0.801075 0.885866 0.822222 0.910603
## 14 0.786096 0.868780 0.817694 0.905361
## 15 0.790323 0.816990 0.814660 0.895565
## 16 0.679144 0.670164 0.801072 0.872965
## capture_rate cumulative_capture_rate gain cumulative_gain
## 1 0.009371 0.009371 -8.018037 -8.018037
## 2 0.012048 0.021419 18.262524 5.122243
## 3 0.011379 0.032798 11.692384 7.312290
## 4 0.010710 0.043507 10.962368 8.188309
## 5 0.010040 0.053548 -1.447897 6.240565
## 6 0.047523 0.101071 -4.697644 0.800707
## 7 0.052878 0.153949 6.040650 2.541117
## 8 0.054886 0.208835 10.067510 4.417671
## 9 0.104418 0.313253 4.138479 4.324441
## 10 0.103748 0.417001 3.470925 4.110776
## 11 0.106426 0.523427 6.711793 4.629307
## 12 0.095047 0.618474 -4.697644 3.078983
## 13 0.099732 0.718206 0.000360 2.640190
## 14 0.098394 0.816600 -1.869510 2.074967
## 15 0.098394 0.914993 -1.341927 1.696217
## 16 0.085007 1.000000 -15.220597 0.000000
## kolmogorov_smirnov
## 1 -0.004106
## 2 0.005247
## 3 0.011235
## 4 0.016553
## 5 0.015812
## 6 0.004036
## 7 0.019178
## 8 0.044415
## 9 0.065275
## 10 0.082770
## 11 0.116419
## 12 0.092867
## 13 0.092869
## 14 0.083446
## 15 0.076718
## 16 0.000000