[R] Recursive feature elimination keeping the weights constant

[Priyanka Purkayastha]

Dear All,

I am trying to build a model by doing recursive elimination of weights one
by one.

This is the example matrix

ID 885038 885039 885040 885041 885042 885043         Label
weights 0.000236 0.004591 0.00017 0.018113 0.000238 0.006537 N/A
1267359 2 0 0 0 0 1 1
1295720 0 0 0 0 0 1 1
1295721 0 0 0 0 0 1 1
1295723 0 0 0 0 0 1 0
1295724 0 0 0 1 0 1 0
1296724 0 0 0 1 0 1 0
12957243 0 0 0 0 0 1 0
12957424 0 0 0 1 0 1 0
12967244 0 0 0 1 0 1 0
12673529 2 0 0 0 0 1 1
1295720 0 0 0 0 0 1 1
12957221 0 0 0 0 0 1 1
Bellow is the code I have written to eliminate minimum rows of weights one
by one and build SVM model.

library(e1071)
library(caret)
library(gplots)
library(ROCR)

data <- read.csv("data.csv", header = TRUE)
rownames(data) <- data[,1]
data<-data[,-1]

for (k in 1:ncol(data))
  {
  rowMin = which.min(data[1,])
  data = data[-rowMin,]
  data = data[-1,]

  inTraining <- createDataPartition(data$Class, p = .70, list = FALSE)
  training <- data[ inTraining,]
  testing  <- data[-inTraining,]

  ## Building the model ####
  svm.model <- svm(Label ~ ., data = training,
cross=10,metric="ROC",type="eps-regression",kernel="linear",na.action=na.omit,probability
= TRUE)

  #prediction and ROC
  svm.model$index
  svm.pred <- predict(svm.model, testing, probability = TRUE)

  #calculating auc
  c <- as.numeric(svm.pred)
  c = c - 1
  pred <- prediction(c, testing$Label)
  perf <- performance(pred,"tpr","fpr")
  plot(perf,fpr.stop=0.1)
  auc <- performance(pred, measure = "auc")
  auc <- auc using y.values[[1]]
  print(paste(ncol(data),colnames(data)[rowMin],auc))

  }

I want my output, like

number of columns, colname with minimum weight, AUC
5 , 885039, 0.67

But I get the following error
Error in svm.default(x, y, scale = scale, ..., na.action = na.action) :
  ‘cross’ cannot exceed the number of observations!
In addition: Warning message:
In svm.default(x, y, scale = scale, ..., na.action = na.action) :
  Variable(s) ‘X885039’ and ‘X885040’ and ‘X885042’ and ‘X885043’ constant.
Cannot scale data.

I

	[[alternative HTML version deleted]]