R 中的 KNN 混淆矩阵？

Question

How do I produce a confusion matrix to see the precision, recall and F score for a KNN model in R? Here's my R code for the KNN model:

nn <- knn(train = train.norm.df[, 1:14], test = new.norm.df, cl = train.norm.df[, 3], k = 3)

How do I display this model, nn in a confusion matrix?

Markdown file: https://rpubs.com/evanmullen36/840780

Answer 1

You should always provide reproducible data with your question and code. Here is an example of what you are trying to do using the iris data set which is included with R. First we need training and testing subsets of the data:

library(class)
data(iris)

set.seed(42)
idx <- seq(nrow(iris))
train <- unname(unlist(tapply(idx, iris$Species, sample, size=25)))
train <- sort(train)
test <- idx[!idx %in% train]

Now you can use the nearest neighbor function:

predict <- knn(iris[train, 1:4], iris[test, 1:4], iris$Species[train])
(conftbl <- table(Species=iris$Species[test], Predict=predict))
#             Predict
# Species      setosa versicolor virginica
#   setosa         25          0         0
#   versicolor      0         25         0
#   virginica       0          2        23

Correct <- sum(diag(conftbl))/sum(conftbl) * 100
cat("Percent Correct = ", round(Correct, 2))
# Percent Correct =  97.33

Notice that in your example you included a column in both the train= and cl= arguments (column 3 is in both). That means that you used the class membership to predict the class membership which is circular. If column 3 is the class membership you should use train = train.norm.df[, c(1:2, 4:14)] and test = new.norm.df[, c(1:2, 4:14)] .