如何在R中为随机森林设置“分类”类型

Question

I have some training ( train ) and testing ( test ) data. This is the way I'm building a forest:

forest <- randomForest(y ~ . - y, data=train, nodesize=25, ntree=200, type="classification")

I wish to emphasize that train$y and test$y may take only integer values like 1, 2, 3, 4 , etc. But when I build my prediction like

pred = predict(forest, newdata=test, type="class")

And check pred contents with:

pred[1:10]

I see such an output:

3        5       12       14       26       27       33       48       50 
2.863208 2.466447 2.476652 2.894254 4.726897 2.378187 2.034159 3.977969 2.264780  

Whereas I expected to see something like:

3         5        12      14
1         3         1       7    # <- I wish to see integers here

So, what is wrong with that?

Answer 1

Change your code as follows:

forest <- randomForest(factor(y) ~., data=train, nodesize=25, ntree=200)

here using -y in the formula is redundant as has been already mentioned by 'LyzandeR'