R - Caret train() “Error: Stopping” with “Not all variable names used in object found in newdata”

问题

I am trying to build a simple Naive Bayes classifer for mushroom data. I want to use all of the variables as categorical predictors to predict if a mushroom is edible.

I am using caret package.

Here is my code in full:

##################################################################################
# Prepare R and R Studio environment
##################################################################################

# Clear the R studio console
cat("\014")

# Remove objects from environment
rm(list = ls())

# Install and load packages if necessary
if (!require(tidyverse)) {
  install.packages("tidyverse")
  library(tidyverse)
}
if (!require(caret)) {
  install.packages("caret")
  library(caret)
}
if (!require(klaR)) {
  install.packages("klaR")
  library(klaR)
}

#################################

mushrooms <- read.csv("agaricus-lepiota.data", stringsAsFactors = TRUE, header = FALSE)

na.omit(mushrooms)

names(mushrooms) <- c("edibility", "capShape", "capSurface", "cap-color", "bruises", "odor", "gill-attachment", "gill-spacing", "gill-size", "gill-color", "stalk-shape", "stalk-root", "stalk-surface-above-ring", "stalk-surface-below-ring", "stalk-color-above-ring", "stalk-color-below-ring", "veil-type", "veil-color", "ring-number", "ring-type", "spore-print-color", "population", "habitat")

# convert bruises to a logical variable
mushrooms$bruises <- mushrooms$bruises == 't'

set.seed(1234)
split <- createDataPartition(mushrooms$edibility, p = 0.8, list = FALSE)

train <- mushrooms[split, ]
test <- mushrooms[-split, ]

predictors <- names(train)[2:20] #Create response and predictor data

x <- train[,predictors] #predictors
y <- train$edibility #response

train_control <- trainControl(method = "cv", number = 1) # Set up 1 fold cross validation

edibility_mod1 <- train( #train the model
  x = x,
  y = y,
  method = "nb", 
  trControl = train_control
)

When executing the train() function I get the following output:

Something is wrong; all the Accuracy metric values are missing:
    Accuracy       Kappa    
 Min.   : NA   Min.   : NA  
 1st Qu.: NA   1st Qu.: NA  
 Median : NA   Median : NA  
 Mean   :NaN   Mean   :NaN  
 3rd Qu.: NA   3rd Qu.: NA  
 Max.   : NA   Max.   : NA  
 NA's   :2     NA's   :2    
Error: Stopping
In addition: Warning messages:
1: predictions failed for Fold1: usekernel= TRUE, fL=0, adjust=1 Error in predict.NaiveBayes(modelFit, newdata) : 
  Not all variable names used in object found in newdata
 
2: model fit failed for Fold1: usekernel=FALSE, fL=0, adjust=1 Error in x[, 2] : subscript out of bounds
 
3: In nominalTrainWorkflow(x = x, y = y, wts = weights, info = trainInfo,  :
  There were missing values in resampled performance measures.

x and y after script run:

> str(x)
'data.frame':   6500 obs. of  19 variables:
 $ capShape                : Factor w/ 6 levels "b","c","f","k",..: 6 6 1 6 6 6 1 1 6 1 ...
 $ capSurface              : Factor w/ 4 levels "f","g","s","y": 3 3 3 4 3 4 3 4 4 3 ...
 $ cap-color               : Factor w/ 10 levels "b","c","e","g",..: 5 10 9 9 4 10 9 9 9 10 ...
 $ bruises                 : logi  TRUE TRUE TRUE TRUE FALSE TRUE ...
 $ odor                    : Factor w/ 9 levels "a","c","f","l",..: 7 1 4 7 6 1 1 4 7 1 ...
 $ gill-attachment         : Factor w/ 2 levels "a","f": 2 2 2 2 2 2 2 2 2 2 ...
 $ gill-spacing            : Factor w/ 2 levels "c","w": 1 1 1 1 2 1 1 1 1 1 ...
 $ gill-size               : Factor w/ 2 levels "b","n": 2 1 1 2 1 1 1 1 2 1 ...
 $ gill-color              : Factor w/ 12 levels "b","e","g","h",..: 5 5 6 6 5 6 3 6 8 3 ...
 $ stalk-shape             : Factor w/ 2 levels "e","t": 1 1 1 1 2 1 1 1 1 1 ...
 $ stalk-root              : Factor w/ 5 levels "?","b","c","e",..: 4 3 3 4 4 3 3 3 4 3 ...
 $ stalk-surface-above-ring: Factor w/ 4 levels "f","k","s","y": 3 3 3 3 3 3 3 3 3 3 ...
 $ stalk-surface-below-ring: Factor w/ 4 levels "f","k","s","y": 3 3 3 3 3 3 3 3 3 3 ...
 $ stalk-color-above-ring  : Factor w/ 9 levels "b","c","e","g",..: 8 8 8 8 8 8 8 8 8 8 ...
 $ stalk-color-below-ring  : Factor w/ 9 levels "b","c","e","g",..: 8 8 8 8 8 8 8 8 8 8 ...
 $ veil-type               : Factor w/ 1 level "p": 1 1 1 1 1 1 1 1 1 1 ...
 $ veil-color              : Factor w/ 4 levels "n","o","w","y": 3 3 3 3 3 3 3 3 3 3 ...
 $ ring-number             : Factor w/ 3 levels "n","o","t": 2 2 2 2 2 2 2 2 2 2 ...
 $ ring-type               : Factor w/ 5 levels "e","f","l","n",..: 5 5 5 5 1 5 5 5 5 5 ...



> str(y)
 Factor w/ 2 levels "e","p": 2 1 1 2 1 1 1 1 2 1 ...

My environment is:

> R.version
               _                           
platform       x86_64-apple-darwin17.0     
arch           x86_64                      
os             darwin17.0                  
system         x86_64, darwin17.0          
status                                     
major          4                           
minor          0.3                         
year           2020                        
month          10                          
day            10                          
svn rev        79318                       
language       R                           
version.string R version 4.0.3 (2020-10-10)
nickname       Bunny-Wunnies Freak Out     
> RStudio.Version()
$citation

To cite RStudio in publications use:

  RStudio Team (2020). RStudio: Integrated Development Environment for R. RStudio, PBC, Boston, MA URL http://www.rstudio.com/.

A BibTeX entry for LaTeX users is

  @Manual{,
    title = {RStudio: Integrated Development Environment for R},
    author = {{RStudio Team}},
    organization = {RStudio, PBC},
    address = {Boston, MA},
    year = {2020},
    url = {http://www.rstudio.com/},
  }


$mode
[1] "desktop"

$version
[1] ‘1.3.1093’

$release_name
[1] "Apricot Nasturtium"

回答1:

What you are trying to do is a bit tricky, most naive bayes implementation or at least the one you are using (from kLAR which is derived from e1071) uses a normal distribution. You can see under the details of naiveBayes help page from e1071:

The standard naive Bayes classifier (at least this implementation) assumes independence of the predictor variables, and Gaussian distribution (given the target class) of metric predictors. For attributes with missing values, the corresponding table entries are omitted for prediction.

And your predictors are categorical so this might be problematic. You can try to set kernel=TRUE and adjust=1 to force it towards normal, and avoid kernel=FALSE which will throw the error.

Before that we remove columns with only 1 level and sort out the column names, also in this case it's easier to use the formula and avoid the making dummy variables :

df = train 
levels(df[["veil-type"]])
[1] "p"
df[["veil-type"]]=NULL
colnames(df) = gsub("-","_",colnames(df))

Grid = expand.grid(usekernel=TRUE,adjust=1,fL=c(0.2,0.5,0.8))

mod1 <- train(edibility~.,data=df,
  method = "nb", trControl = trainControl(method="cv",number=5),
  tuneGrid=Grid
)

 mod1
Naive Bayes 

6500 samples
  21 predictor
   2 classes: 'e', 'p' 

No pre-processing
Resampling: Cross-Validated (5 fold) 
Summary of sample sizes: 5200, 5200, 5200, 5200, 5200 
Resampling results across tuning parameters:

  fL   Accuracy   Kappa    
  0.2  0.9243077  0.8478624
  0.5  0.9243077  0.8478624
  0.8  0.9243077  0.8478624

Tuning parameter 'usekernel' was held constant at a value of TRUE

Tuning parameter 'adjust' was held constant at a value of 1
Accuracy was used to select the optimal model using the largest value.
The final values used for the model were fL = 0.2, usekernel = TRUE and
 adjust = 1.

来源：https://stackoverflow.com/questions/64757800/r-caret-train-error-stopping-with-not-all-variable-names-used-in-object

标签

machine-learning

r-caret

naivebayes