Calculating accuracy of prediction of rpart model

Question

I have this modified iris data-set which comprises of first 100 rows i.e only the 'setosa' and 'versicolor' species. I have randomized the rows using sample() function:

iris1[sample(nrow(iris1)),]->iris1

The i've divided the data-set in 65:35 ratio using the sample.split function from caTools package:

sample.split(iris1$Species,SplitRatio = 0.65)->mysplit
subset(iris1,mysplit==T)->train

subset(iris1,mysplit==F)->test

Following which i've built the rpart model on top of "train" set and predicted the values on "test" set:

rpart(Species~.,data=train)->mod1

predict(mod1,test,type = "class")->result1

Now, i would want to find the accuracy of prediction on the test set, how can i do that?

Answer 1

Your first task would be to build a confusion matrix for the actual values and predicted value, you can do that using the table() function:

table(test$Species,result1)

This would give you the below confusion matrix:

             result1
             setosa versicolor
  setosa         18          0
  versicolor      0         18

Now, you can find out the accuracy of prediction by dividing the correctly predicted results upon all the results:

(18+18)/(18+18+0+0)

This would give you an accuracy of 100%

You can also use the "confusionMatrix()" function from the caret package:

confusionMatrix(table(test$Species,result1))

This would be the result:

Accuracy : 1         
                 95% CI : (0.9026, 1)
    No Information Rate : 0.5       
    P-Value [Acc > NIR] : 1.455e-11