Every supervised learning problem falls into one of two kinds. Regression predicts a number, such as a price, a temperature, or days until delivery. Classification predicts a category, such as spam or not spam, churn or stay, pass or fail. Choosing the wrong kind leads to a poor result, because the model optimises for the wrong thing.
Telling them apart in practice
- Look at the target column: continuous numbers mean regression, while a small set of labels means classification.
- Check the error measure: regression uses errors in units, while classification uses accuracy, precision, and recall.
- Watch for disguised classification, such as predicting a rating from one to five, which is often treated as categories.
- Remember that some problems can be framed either way, and the framing should follow how the answer will be used.
In interviews, state the problem type before naming any algorithm. It shows you think about the problem first and the tool second.