Question

我正在尝试使用caret包编写决策树模型，但无法使其正常工作。

首先，我想看看该模型是否与rpart包一起工作，在这里我可以运行相同的模型，并且可以运行-

# setup
set.seed(123)
library(rpart)
library(caret)

# reading the file containing spam data
spamD <- readr::read_tsv(
  "https://raw.githubusercontent.com/WinVector/zmPDSwR/master/Spambase/spamD.tsv"
)
#> Parsed with column specification:
#> cols(
#>   .default = col_double(),
#>   spam = col_character()
#> )
#> See spec(...) for full column specifications.

# creating training and testing datasets
spamTrain <- dplyr::filter(.data = spamD, rgroup >= 10)
spamTest <- dplyr::filter(.data = spamD, rgroup < 10)

# training the model (works)
(treemodel <- rpart::rpart(formula = spam == "spam" ~ ., 
                          data = dplyr::select(spamTrain, -rgroup)))
#> n= 4143 
#> 
#> node), split, n, deviance, yval
#>       * denotes terminal node
#> 
#>  1) root 4143 989.338600 0.39415880  
#>    2) char.freq.bang< 0.0795 2381 308.352800 0.15287690  
#>      4) word.freq.remove< 0.045 2210 199.699500 0.10045250  
#>        8) char.freq.dollar< 0.164 2138 156.482700 0.07951356  
#>         16) word.freq.free< 0.115 1968 110.044200 0.05945122 *
#>         17) word.freq.free>=0.115 170  36.476470 0.31176470 *
#>        9) char.freq.dollar>=0.164 72  14.444440 0.72222220 *
#>      5) word.freq.remove>=0.045 171  24.081870 0.83040940  
#>       10) word.freq.george>=0.08 14   0.000000 0.00000000 *
#>       11) word.freq.george< 0.08 157  13.566880 0.90445860 *
#>    3) char.freq.bang>=0.0795 1762 355.060700 0.72020430  
#>      6) capital.run.length.average< 2.3995 625 150.198400 0.40160000  
#>       12) word.freq.free< 0.075 454  85.374450 0.25110130  
#>         24) word.freq.remove< 0.045 409  60.611250 0.18092910  
#>           48) word.freq.internet< 0.08 377  43.368700 0.13262600 *
#>           49) word.freq.internet>=0.08 32   6.000000 0.75000000 *
#>         25) word.freq.remove>=0.045 45   4.444444 0.88888890 *
#>       13) word.freq.free>=0.075 171  27.239770 0.80116960 *
#>      7) capital.run.length.average>=2.3995 1137 106.545300 0.89533860  
#>       14) word.freq.hp>=0.41 51   6.745098 0.15686270 *
#>       15) word.freq.hp< 0.41 1086  70.681400 0.93001840  
#>         30) word.freq.edu>=0.52 15   0.000000 0.00000000 *
#>         31) word.freq.edu< 0.52 1071  57.525680 0.94304390 *

但是，如果我使用caret包运行相同的模型，则该模型不会运行-

# using `caret` package to do the same (doesn't work)
caret::train(
  formula = spam == "spam" ~ ., 
  data = dplyr::select(spamTrain, -rgroup),
  method = "rpart"
)

#> Something is wrong; all the RMSE metric values are missing:
#>       RMSE        Rsquared        MAE     
#> Min.   : NA   Min.   : NA   Min.   : NA  
#> 1st Qu.: NA   1st Qu.: NA   1st Qu.: NA  
#> Median : NA   Median : NA   Median : NA  
#> Mean   :NaN   Mean   :NaN   Mean   :NaN  
#> 3rd Qu.: NA   3rd Qu.: NA   3rd Qu.: NA  
#> Max.   : NA   Max.   : NA   Max.   : NA  
#> NA's   :3     NA's   :3     NA's   :3    
#> Error: Stopping
#> In addition: There were 26 warnings (use warnings() to see them)

Answer 1

从?caret::train中可以看到，没有formula参数，而是形式。另外，您需要重新格式化并过滤NAs：

caret::train(
  form= spam ~  ., 
  data = (dplyr::select(spamTrain, -rgroup) %>% filter(!is.na(word.freq.cs))),
  method = "rpart"
)

最好！

使用`caret`软件包运行决策树模型时出错

1 个答案: