Confusion matrix in waffles?

[Nobody/Anonymous]

Hi,

Is it possible to get a confusion matrix out of the testing of a model in
waffles? I'd like to know numbers of false positives, false negatives, true
positives and true negatives. Is there any way to do this?

Robin

[Mike Gashler]

Thanks for pointing this out. I just added this feature. So, you'll need to
get the latest source from our Subversion repository (http://waffles.sourcefo
rge.net/tutorial/subversion.html). Here's an example of using the new feature:

waffles_learn train iris.arff naivebayes > model.twt

waffles_learn test -confusion model.twt iris.arff

0.97333333333333

(Rows=expected values, Cols=predicted values, Elements=number of occurrences)

Confusion matrix for class Iris-setosaIris-versicoIris-virgini

Iris-setosa 50 0 0

Iris-versicolor 0 48 2

Iris-virginica 0 2 48

[Nobody/Anonymous]

how can get confusion matrix within program?

[Mike Gashler]

When you call GSupervisedLearner::accuracy, there is an optional parameter
named "pNominalLabelStats". If you pass a pointer to an empty
"vector<gmatrix*>" with this parameter, then it will return a confusion matrix
for each label attribute.</gmatrix*>

[Nobody/Anonymous]

if you want see confusion matrix within program u must write a code like:

'

vector<gmatrix*> confusionMatrices;</gmatrix*>

double* pp;

GTEMPBUF(double, results,punLables->cols());

bool confusion = false;

modell.accuracy(punFeatures, punLables, results, &confusionMatrices);

GVec::print(cout, 14, results,punLables->cols());

cout << "\n";

printConfusionMatrices(punLables->relation().get(), confusionMatrices);