r - Memory usage of caret train function with cforest model -

i trying train() cforest model using caret package , 32 gb of ram being hoovered @ end of process (at 92% seen on screeshot below). cannot share data, question more cforest() implementation (or maybe theory behind it?) actual code. hope still fits here.

data , code:

dataset not big, 230k obs. 19 variables (1 of target), str() follows:

 classes ‘data.table’ , 'data.frame':  227358 obs. of  20 variables:  $ var_n : factor w/ 11 levels  $ var_n : factor w/ 2 levels  $ var_n : factor w/ 2 levels  $ var_n : factor w/ 2 levels  $ var_n : factor w/ 2 levels  $ var_n : factor w/ 2 levels  $ var_n : factor w/ 16 levels  $ var_n : factor w/ 2 levels  $ var_n : num  $ var_n : factor  $ var_n : factor  $ var_n : int  $ var_n : int  $ var_n : factor w/ 15 levels  $ var_n : factor w/ 16 levels  $ var_n : int  $ var_n : factor w/ 7 levels  $ var_n : factor w/ 31 levels 

my whole control , train looks follows:

ctrl_params <- traincontrol(   summaryfunction = twoclasssummary,   # selectionfunction = 'onese',   selectionfunction = 'best',   index = cv_ids_caret,   method = "oob",   # method = "cv", number = folds,   verboseiter = true,   classprobs = true )  set.seed(2016) cforest_caret_model <- train(   x = x_train,   y = y_train,  # factor(y_train),   method = 'cforest',   metric = 'roc',   trcontrol = ctrl_params,   controls = cforest_unbiased(ntree = 11, trace = true) ) 

my sessioninfo:

r version 3.3.2 (2016-10-31) platform: x86_64-w64-mingw32/x64 (64-bit) running under: windows 7 x64 (build 7601) service pack 1     attached base packages: [1] stats4    grid      stats     [4] graphics  grdevices utils     [7] datasets  methods   base       other attached packages:  [1] randomforest_4.6-12  [2] mdmisc_0.0.0.9002    [3] glmnet_2.0-5         [4] foreach_1.4.3        [5] matrix_1.2-8         [6] rpart_4.1-10         [7] party_1.2-1          [8] strucchange_1.5-1    [9] sandwich_2.3-4      [10] zoo_1.7-14          [11] modeltools_0.2-21   [12] mvtnorm_1.0-5       [13] caret_6.0-73        [14] hmisc_4.0-2         [15] ggplot2_2.2.1       [16] formula_1.2-1       [17] survival_2.40-1     [18] lattice_0.20-34     [19] magrittr_1.5        [20] lubridate_1.6.0     [21] stringr_1.1.0       [22] data.table_1.10.4   [23] tidyr_0.6.1         [24] dplyr_0.5.0         

what have tried:

so can see have tried to:

• use oob instead of cv, suggested here;
• limit number of trees ridiculously small number (11); and
• use 'best' model selection instead of 'onese' hoping smaller memory usage,

but still during training see that:

after seeing memory error:

the training continues right maximum:

summary:

i not fixed on using model, conditional inference trees variable selection process (not biased towards variables more categories etc.), understand issue wander bit conditional inference forest.