CONFMAT

Construct confusion matrix

   [C,NE,LABLIST1,LABLIST2] = CONFMAT(LAB1,LAB2,METHOD,FID)

Input
LAB1	Set of labels
LAB2	Set of labels
METHOD	'count' (default) to count number of co-occurences in
LAB1	and LAB2, 'disagreement' to count relative  non-co-occurrence.
FID	Write text result to file

Output
C	Confusion matrix
NE	Total number of errors (empty labels are neglected)
LABLIST1	Label list for LAB1
LABLIST2	Label list for LAB2

Description

Constructs a confusion matrix C between two sets of labels LAB1 (corresponding to the rows in C) and LAB2 (the columns in C). The order of  the rows and columns is returned in LABLIST. NE is the total number of  errors (sum of non-diagonal elements in C).

If METHOD = 'count' (default), co-occurences in LAB1 and LAB2 are counted  and returned in C.  For METHOD = 'disagreement', the relative disagreement is returned in NE,  and is split over all combinations of labels in C (such that the rows sum  to 1). (The total disagreement for a class equals one minus the  sensitivity for that class as computed by TESTC).  For METHOD = 'ids' a cell array C is returned in which C(i,j) contains  the indices of the objects for which LAB1 equals LABLIST1(i,:) and LAB2 equals LABLIST2(j,:).

    [C,NE,LABLIST] = CONFMAT(D,METHOD)

If D is a classification result D = A*W, the labels LAB1 and LAB2 are  internally retrieved by CONFMAT before computing the confusion matrix.

    C = CONFMAT(D)

This call also applies in case in D = A*W the dataset A has soft labels  W is trained by a soft labeld classifier.

When no output argument is specified, or when FID is given, the  confusion matrix is displayed or written a to a text file. It is assumed  that LAB1 contains true labels and LAB2 stores estimated labels.

Example(s)

 Typical use of CONFMAT is the comparison of true and and estimated labels
 of a testset A by application to a trained classifier W: 
 LAB1 = GETLABELS(A); LAB2 = A*W*LABELD.
 More examples can be found in PREX_CONFMAT, PREX_MATCHLAB.

See also

mappings, datasets, getlabels, labeld,