False-positive and false-negative

cyckuan

Have taken a few close looks at the performance vectors coming out of RapidMiner's "Performance (Binominal)" operator.

Could it be possible that the numbers for false positive and false negative be the other way round?

When testing a number of different classifiers on the same data, the actual true or 'true true" totals should remain the same. The only thing that should change is how many with the "true true" column is predicted e.g.
predicted true && true true = true positive
predicted false && true true = false negative

The definitions are referring to the right cells, but the numbers within the cell seems swapped.

cyckuan

Ok my bad. I swapped my prediction and label roles around and everything is falling into the right places again.

All good now.