I have a quick question,
Does accuracy work well for regression problems?
I got zero accuracy although I had very good testing results.
Accuracy is a classification measure. It does not make sense in the context of regressions.
The main reason for that is that a regression is defined by continuous outcomes, while classification with categorical outcomes (classes).
When you have categories, predicting 1 and having an actual observed value of 1 means that we’ve got a match (prediction is accurate).
When we’ve got continuous outcomes though, predicting 0.99999 and having an actual observed value of 1 are two different things. That is why we don’t use the accuracy measure for regressions.