R plot for showing true/false positives

Question

I'm looking for a way to plot my data. Suppose you have a number of algorithms that produce a ranked list, and a classifier that identifies each item in the list as true or false. How can I produce a chart that puts each algorithm beside each other and each rank has its own square stacked as follow:

      __   __
 3   |TP| |FP|
     |__| |__|
 2   |FP| |TP|
     |__| |__|
 1   |TP| |TP|
     |__| |__|
      a    b

Where TP could be coloured green and FP could be coloured red for example.

Maybe a mosiac plot [(LINK)](http://www.statmethods.net/advgraphs/mosaic.html) — Tyler Rinker, Nov 02 '12 at 20:07
Hi @Robert, you may want to check if the solutions solve your issue and accept the best one — HAVB, May 28 '17 at 21:09

score 1 · Answer 1 · edited May 23 '17 at 12:09

1

You may use ggplot2 for drawing a stacked barplot. However a better way to compare the true/false positive/negatives of algorithms is precision/recall (same as selectivity/specificity) chart.

edited May 23 '17 at 12:09

Community

1
1

answered Nov 02 '12 at 20:17

Ali

9,440
12
62
92

score 1 · Answer 2 · answered Apr 18 '17 at 16:04

It can be done very easily with base R, using the heatmap() function.

Try this:

# Dataframe with TRUE/FALSE items
mydf <- data.frame(a = c(T, F, T), b = c(T, T,F))

# Transform booleans into 1s and 0s, and plot heatmap
heatmap(apply(mydf, 2, as.integer),
        col = c("red", "lightgreen"),
        xlab = "Algorithm", 
        ylab = "Item", 
        main = "False positives in red", 
        Rowv = NA, 
        Colv = NA,
        scale = "column")

R plot for showing true/false positives

2 Answers2