问题
I'm trying to put some plots side by side using the cowplot package in R.
Here's the (quite bad) result I am getting:
As you can see, this image is full of troubles: the plots are tiny and too small, the text font sizes are too big, and there's too much blank space left.
I am able to generate and save correctly each of the three plots, but I get that disaster image when I run this command:
general_plot <- plot_grid(plot_cf, plot_pos_neg, plot_scores, ncol = 2, align="h", labels=c("A", "B", "C"), label_size=5)
What I'm I doing wrong?
Here's my complete code (ready to run):
#!/usr/bin/env Rscript
if (!require("RColorBrewer")) {
install.packages("RColorBrewer")
library(RColorBrewer)
}
library(grid)
library(gridExtra)
library(gtable)
library(ggplot2)
library(cowplot)
SAVE_FILES = FALSE
digits = 2
generalTextSize = 5
# Confusion matrix barplot
barplot_creator <- function(inputText, saveFileName, upperYlim) {
confusionMatrixDataFrame <- read.table(header=TRUE, text=inputText)
confusionMatrixDataFrame$category = gsub("_", " ", confusionMatrixDataFrame$category)
confusionMatrixDataFrame$category = factor(confusionMatrixDataFrame$category, levels=unique(confusionMatrixDataFrame$category))
p <- ggplot(confusionMatrixDataFrame, aes(x=category, y=amount)) + geom_bar(aes(fill=category), stat="identity", position=position_dodge()) + ylim(0,upperYlim)
p + scale_colour_brewer(palette="RdBu") + xlab("") + ylab("")
p + theme(legend.text=element_text(size=generalTextSize), text = element_text(size=generalTextSize), axis.text.x = element_text(size=generalTextSize), axis.text.y = element_text(size=generalTextSize))
p
if (SAVE_FILES==TRUE) {
ggsave(saveFileName)
}
return(p)
}
# Three elements barplot
accuracy_f1score_mcc_barplot <- function(accuracy, f1_score, normMCC, saveFileName) {
thisText <- paste("\n category amount\n accuracy=",accuracy, " ", accuracy,"\n F1_score=", f1_score, " ", f1_score,"\n\n normMCC=", normMCC," ", normMCC,"\n", sep="")
upperYlim = 1.0
return(barplot_creator(thisText, saveFileName, upperYlim))
}
# Confusion matrix barplot
confusion_matrix_barplot <- function(tp, fn, tn, fp, saveFileName) {
thisText <- paste("\n category amount\n TP=",tp," ", tp,"\n FN=", fn," ", fn,"\n\n TN=",tn," ", tn,"\n FP=", fp," ", fp,"\n", sep="")
# confusionMatrixDataFrame <- read.table(header=TRUE, text='
# category amount
# TP 47
# FN 3
#
# TN 5
# FP 45
# ')
upperYlim = 100
return(barplot_creator(thisText, saveFileName, upperYlim))
}
# Positives negatives barplot
positives_negatives_barplot <- function(pos, neg, saveFileName) {
thisText <- paste("\n category amount\n positives=",pos, " ", pos,"\n negatives=", neg, " ", neg,"\n", sep="")
upperYlim = 100
return(barplot_creator(thisText, saveFileName, upperYlim))
}
# accuracy
accuracy <- function (tp, fn, tn, fp) {
accuracy_result <- (tp+tn)/(tn + tp + fp + fn)
accuracy_result <- round(accuracy_result, digits)
cat("accuracy = ", round(accuracy_result, digits), "\t\t in the [0; 1] interval \n")
return(accuracy_result)
}
# F1 score
f1_score <- function (tp, fn, tn, fp) {
f1_score_result <- (2*tp)/(2*tp + fp + fn)
f1_score_result <- round(f1_score_result, digits)
cat("F1 score = ", round(f1_score_result, digits), "\t\t in the [0; 1] interval\n")
return(f1_score_result)
}
# Matthews correlation coefficient
mcc <- function (tp, fn, tn, fp) {
TP <- tp
FN <- fn
TN <- tn
FP <- fp
#TP;TN;FP;FN # for debugging
sum1 <- TP+FP; sum2 <-TP+FN ; sum3 <-TN+FP ; sum4 <- TN+FN;
denom <- as.double(sum1)*sum2*sum3*sum4 # as.double to avoid overflow error on large products
if (any(sum1==0, sum2==0, sum3==0, sum4==0)) {
denom <- 1
}
mcc <- ((TP*TN)-(FP*FN)) / sqrt(denom)
mcc <- round(mcc, digits)
cat("MCC = ", round(mcc, digits), "\t\t\t in the [-1; +1] interval\n")
return(mcc)
}
# Confusion matrix pie
confusion_matrix_pie <- function(tp, fn, tn, fp, saveFileName) {
thisDataFrame <- data.frame(
category = c(paste("TP=", tp, ""), paste("FN=", fn, ""), paste("TN=", tn, ""), paste("FP=", fp, "")),
amount = c(tp, fn, tn, fp)
)
thisDataFrame$category = factor(thisDataFrame$category, levels=unique(thisDataFrame$category)) # we set as levels to respect this order in the plot legend
head(thisDataFrame)
# Barplot
thisBarPlot<- ggplot(thisDataFrame, aes(x="", y=amount, fill=category))+
geom_bar(width = 1.0, stat = "identity")
thisPie <- thisBarPlot + coord_polar("y") + xlab("") + ylab("") + scale_fill_brewer(palette="RdBu")
thisPie + theme(legend.text=element_text(size=generalTextSize))
thisPie
if (SAVE_FILES==TRUE) {
ggsave(saveFileName)
}
return(thisPie)
}
# positive_negative pie
positive_negative_pie <- function(pos, neg, saveFileName) {
thisDataFrame <- data.frame(
category = c(paste("positives=", pos, ""), paste("negatives=", neg, "")),
amount = c(pos, neg)
)
thisDataFrame$category = factor(thisDataFrame$category, levels=unique(thisDataFrame$category)) # we set as levels to respect this order in the plot legend
head(thisDataFrame)
# Barplot
thisBarPlot<- ggplot(thisDataFrame, aes(x="", y=amount, fill=category))+
geom_bar(width = 1.0, stat = "identity")
thisPie <- thisBarPlot + coord_polar("y") + xlab("") + ylab("") + scale_fill_brewer(palette="RdBu")
thisPie + theme(legend.text=element_text(size=generalTextSize))
thisPie
if (SAVE_FILES==TRUE) {
ggsave(saveFileName)
}
return(thisPie)
}
#
# Function that generates all the plot and their files
#
generate_all_the_plots <- function(tp, fn, tn, fp, addTitle) {
randomValue = sample(1:10000000, 1)
test_title = paste(addTitle, "_test",randomValue,"__TP",tp,"_FN",fn,"_TN",tn,"_FP",fp, sep="")
positives = tp + fn
negatives = tn + fp
confusion_matrix_pie_file_name = paste("../plots/", test_title, "_confusion_matrix_pie.pdf", sep="")
plot_cf <- confusion_matrix_pie(tp, fn, tn, fp, confusion_matrix_pie_file_name)
positive_negative_pie_file_name = paste("../plots/", test_title, "_positive_negative_pie.pdf", sep="")
plot_pos_neg<-positive_negative_pie(positives, negatives, positive_negative_pie_file_name)
current_accuracy = accuracy(tp, fn, tn, fp)
current_f1_score = f1_score(tp, fn, tn, fp)
current_MCC = mcc(tp, fn, tn, fp)
current_normMCC = (current_MCC+1)/2 # normalized MCC
current_normMCC = round(current_normMCC, digits)
cat("normMCC = ", round(current_normMCC, digits), "\t\t in the [0; 1] interval\n")
confusion_matrix_barplot_file_name = paste("../plots/", test_title, "_confusion_matrix_barplot.pdf", sep="")
# confusion_matrix_barplot(tp, fn, tn, fp, confusion_matrix_barplot_file_name)
positives_negatives_barplot_file_name = paste("../plots/", test_title, "_positives_negatives_barplot.pdf", sep="")
# positives_negatives_barplot(positives, negatives, positives_negatives_barplot_file_name)
accuracy_f1score_mcc_barplot_file_name = paste("../plots/", test_title, "_accuracy_f1score_mcc.pdf", sep="")
plot_scores <- accuracy_f1score_mcc_barplot(current_accuracy, current_f1_score, current_normMCC, accuracy_f1score_mcc_barplot_file_name)
## Side by side
# right_col_plot = plot_grid(plot_cf, plot_pos_neg, labels=c("b", "c"), align="v", ncol=1)
# plot_grid(plot_scores, right_col_plot, ncol = 2, rel_widths=c(1.3, 1))
theme_set(theme_cowplot(font_size=12))
general_plot <- plot_grid(plot_cf, plot_pos_neg, plot_scores, ncol = 2, align="h", labels=c("A", "B", "C"), label_size=5)
save_plot("general.pdf", general_plot)
#(plot_cf, plot_pos_neg, plot_scores)
}
exampleA = c(90, 1, 0, 9)
exampleB = c(47, 3, 5, 45)
exampleC = c(1, 9, 89, 1)
# tp = exampleC[1]
# fn = exampleC[2]
# tn = exampleC[3]
# fp = exampleC[4]
selected_example <- exampleC
addTitle = "ExampleC"
#library(gsubfn) # need 0.7-0 or later
generate_all_the_plots(selected_example[1], selected_example[2], selected_example[3], selected_example[4], addTitle)
#if (file.exists("../plots/Rplots.pdf")) file.remove("../plots/Rplots.pdf")
The image generated is saved into "general.pdf". Can anyone tell me what I'm doing wrong?
Thanks!
回答1:
Replacing the save_plot() line by this:
save_plot("general.pdf", general_plot, ncol = 2, nrow = 2)
I get this result, which seems mostly fine:
The labels (A, B, C) are way too small, but that's because you set their size to 5. It should be at least 12. And the axis tick labels in part c overlap, but that's unavoidable for those kinds of plots. I would recommend flipping the axes so the bars run horizontally.
来源：https://stackoverflow.com/questions/49440858/r-why-is-cowplot-plot-grid-generating-tiny-pictures-side-by-side