ggplot2 confusion matrix geom_text labeling - r

Question

I've plotted a confusion matrix (predicting 5 outcomes) in R using ggplot and scales for geom_text labeling.
The way geom_text(aes(label = percent(Freq/sum(Freq))) is written in code, it's showing Frequency of each box divided by sum of all observations, but what I want to do is get Frequency of each box divided by sum Frequency for each Reference.
In other words, instead of A,A = 15.8%,
it should be A,A = 15.8%/(0.0%+0.0%+0.0%+0.0%+15.8%%) = 100.0%
library(ggplot2)
library(scales)
valid_actual <- as.factor(c("A","B","B","C","C","C","E","E","D","D","A","A","A","E","E","D","D","C","B"))
valid_pred <- as.factor(c("A","B","C","C","E","C","E","E","D","B","A","B","A","E","D","E","D","C","B"))
cfm <- confusionMatrix(valid_actual, valid_pred)
ggplotConfusionMatrix <- function(m){
mytitle <- paste("Accuracy", percent_format()(m$overall[1]),
"Kappa", percent_format()(m$overall[2]))
p <-
ggplot(data = as.data.frame(m$table) ,
aes(x = Reference, y = Prediction)) +
geom_tile(aes(fill = log(Freq)), colour = "white") +
scale_fill_gradient(low = "white", high = "green") +
geom_text(aes(x = Reference, y = Prediction, label = percent(Freq/sum(Freq)))) +
theme(legend.position = "none") +
ggtitle(mytitle)
return(p)
}
ggplotConfusionMatrix(cfm)

Answer 1

The problem is that, as far as I know, ggplot is not able to do group calculation. See this recent post for similar question.
To solve your problem you should take advantage of the dplyrpackage.
This should work
library(ggplot2)
library(scales)
library(caret)
library(dplyr)
valid_actual <- as.factor(c("A","B","B","C","C","C","E","E","D","D","A","A","A","E","E","D","D","C","B"))
valid_pred <- as.factor(c("A","B","C","C","E","C","E","E","D","B","A","B","A","E","D","E","D","C","B"))
cfm <- confusionMatrix(valid_actual, valid_pred)
ggplotConfusionMatrix <- function(m){
mytitle <- paste("Accuracy", percent_format()(m$overall[1]),
"Kappa", percent_format()(m$overall[2]))
data_c <- mutate(group_by(as.data.frame(m$table), Reference ), percentage =
percent(Freq/sum(Freq)))
p <-
ggplot(data = data_c,
aes(x = Reference, y = Prediction)) +
geom_tile(aes(fill = log(Freq)), colour = "white") +
scale_fill_gradient(low = "white", high = "green") +
geom_text(aes(x = Reference, y = Prediction, label = percentage)) +
theme(legend.position = "none") +
ggtitle(mytitle)
return(p)
}
ggplotConfusionMatrix(cfm)
And the result: