Confusion matrices compare two classifications (usually one done automatically
using a machine learning algorithm versus the true classication represented by
a manual classification by a specialist... but one can also compare two
automatic or two manual classifications against each other).
Usage
confusion(x, ...)
## Default S3 method:
confusion(x, y = NULL, vars = c("Actual", "Predicted"),
labels = vars, merge.by = "Id", useNA = "ifany", prior, ...)
## S3 method for class 'mlearning'
confusion(x, y = response(x),
labels = c("Actual", "Predicted"), useNA = "ifany", prior, ...)
## S3 method for class 'confusion'
print(x, sums = TRUE, error.col = sums, digits = 0,
sort = "ward", ...)
## S3 method for class 'confusion'
summary(object, type = "all", sort.by = "Fscore",
decreasing = TRUE, ...)
## S3 method for class 'summary.confusion'
print(x, ...)
## S3 method for class 'confusion'
plot(x, y = NULL, type = c("image", "barplot", "stars",
"dendrogram"), stat1 = "Recall", stat2 = "Precision", names, ...)
confusionImage(x, y = NULL, labels = names(dimnames(x)), sort = "ward",
numbers = TRUE, digits = 0, mar = c(3.1, 10.1, 3.1, 3.1), cex = 1, asp = 1,
colfun, ncols = 41, col0 = FALSE, grid.col = "gray", ...)
confusionBarplot(x, y = NULL, col = c("PeachPuff2", "green3", "lemonChiffon2"),
mar = c(1.1, 8.1, 4.1, 2.1), cex = 1, cex.axis = cex, cex.legend = cex,
main = "F-score (precision versus recall)", numbers = TRUE, min.width = 17,
...)
confusionStars(x, y = NULL, stat1 = "Recall", stat2 = "Precision", names, main,
col = c("green2", "blue2", "green4", "blue4"), ...)
confusionDendrogram(x, y = NULL, labels = rownames(x), sort = "ward",
main = "Groups clustering", ...)
prior(object, ...)
## S3 method for class 'confusion'
prior(object, ...)
prior(object, ...) <- value
## S3 replacement method for class 'confusion'
prior(object, ...) <- value
Arguments
x
an object.
y
another object, from which to extract the second classification, or
NULL if not used.
vars
the variables of interest in the first and second classification
in the case the objects are lists or data frames. Otherwise, this argument
is ignored and x and y must be factors with same length and
same levels.
labels
labels to use for the two classifications. By default, it is
the same as vars or the one in the confusion matrix.
merge.by
a character string with the name of variables to use to merge
the two data frames, or NULL.
useNA
do we keep NAs as a separate category? The default "ifany"
creates this category only if there are missing values. Other possibilities
are "no", or "always".
prior
class frequencies to use for first classifier that
is tabulated in the rows of the confusion matrix. For its value, see here
under, the value argument.
sums
is the confusion matrix printed with rows and columns sums?
error.col
is a column with class error for first classifier added
(equivalent to flase negative rate of FNR)?
digits
the number of digits after the decimal point to print in the
confusion matrix. The default or zero leads to most compact presentation
and is suitable for frequencies, but not for relative frequencies.
sort
are rows and columns of the confusion matrix sorted so that
classes with larger confusion are closer together? Sorting is done
using a hierachical clustering with hclust(). The clustering method
is provided is the one provides ("ward", by default, but see the
hclust() help for other options). If FALSE or NULL, no
sorting is done.
object
a 'confusion' object.
sort.by
the statistics to use to sort the table (by default, Fmeasure,
the F1 score for each class = 2 * recall * precision / (recall + precision)).
decreasing
do we sort in increasing or decreasing order?
type
the type of graph to plot (only "stars" if two confusion
matrices are to be compared).
stat1
first statistic to compare in the stars plot.
stat2
second statistic to compare in the stars plot.
...
further arguments passed to the function. In particular for
plot(), it can be all arguments for the corresponding plot.
numbers
are actual numbers indicated in the confusion matrix image?
mar
graph margins.
cex
text magnification factor.
cex.axis
idem for axes. If NULL, the axis is not drawn.
cex.legend
idem for legend text. If NULL, no legend is added.
asp
graph aspect ration. There is little reasons to cvhange the
default value of 1.
col
color(s) to use fir the graph.
colfun
a function that calculates a series of colors, like e.g.,
cm.colors() and that accepts one argument being the number of colors
to be generated.
ncols
the number of colors to generate. It should preferrably be
2 * number of levels + 1, where levels is the number of frequencies you
want to evidence in the plot. Default to 41.
col0
should null values be colored or not (no, by default)?
grid.col
color to use for grid lines, or NULL for not drawing
grid lines.
names
names of the two classifiers to compare.
main
main title of the graph.
min.width
minimum bar width required to add numbers.
value
a single positive numeric to set all class frequencies to this
value (use 1 for relative frequencies and 100 for relative freqs in percent),
or a vector of positive numbers of the same length as the levels in the
object. If the vector is named, names must match levels. Alternatively,
providing NULL or an object of null length resets row class
frequencies into their initial values.
Value
A confusion matrix in a 'confusion' object. prior() returns the
current class frequencies associated with first classification tabulated, i.e.,
for rows in the confusion matrix.
Author(s)
Philippe Grosjean <Philippe.Grosjean@umons.ac.be> and
Kevin Denis <Kevin.Denis@umons.ac.be>
See Also
mlearning, hclust,
cm.colors
Examples
data("Glass", package = "mlbench")
## Use a little bit more informative labels for Type
Glass$Type <- as.factor(paste("Glass", Glass$Type))
## Use learning vector quantization to classify the glass types
## (using default parameters)
summary(glassLvq <- mlLvq(Type ~ ., data = Glass))
## Calculate cross-validated confusion matrix and plot it in different ways
(glassConf <- confusion(cvpredict(glassLvq), Glass$Type))
## Raw confusion matrix: no sort and no margins
print(glassConf, sums = FALSE, sort = FALSE)
## Graphs
plot(glassConf) # Image by default
plot(glassConf, sort = FALSE) # No sorting
plot(glassConf, type = "barplot")
plot(glassConf, type = "stars")
plot(glassConf, type = "dendrogram")
summary(glassConf)
summary(glassConf, type = "Fscore")
## Build another classifier and make a comparison
summary(glassNaiveBayes <- mlNaiveBayes(Type ~ ., data = Glass))
(glassConf2 <- confusion(cvpredict(glassNaiveBayes), Glass$Type))
## Comparison plot for two classifiers
plot(glassConf, glassConf2)
## When the probabilities in each class do not match the proportions in the
## training set, all these calculations are useless. Having an idea of
## the real proportions (so-called, priors), one should first reweight the
## confusion matrix before calculating statistics, for instance:
prior1 <- c(10, 10, 10, 100, 100, 100) # Glass types 1-3 are rare
prior(glassConf) <- prior1
glassConf
summary(glassConf, type = c("Fscore", "Recall", "Precision"))
plot(glassConf)
## This is very different than if glass types 1-3 are abundants!
prior2 <- c(100, 100, 100, 10, 10, 10) # Glass types 1-3 are abundants
prior(glassConf) <- prior2
glassConf
summary(glassConf, type = c("Fscore", "Recall", "Precision"))
plot(glassConf)
## Weight can also be used to construct a matrix of relative frequencies
## In this case, all rows sum to one
prior(glassConf) <- 1
print(glassConf, digits = 2)
## However, it is easier to work with relative frequencies in percent
## and one gets a more compact presentation
prior(glassConf) <- 100
glassConf
## To reset row class frequencies to original propotions, just assign NULL
prior(glassConf) <- NULL
glassConf
prior(glassConf)