At the risk of generalizing too much, let's try to summarize a sample decision problem for a model:
- Are we trying to characterize data by simply grouping information based on its characteristics, without any or a few previous hints? This is the domain of clustering techniques.
- The first and most basic question: are we trying to predict the instant outcome of a variable, or to tag or classify data into groups? If the former, we are tackling a regression problem. If the latter, this is the realm of classification problems.
- Having the former questions resolved, and opting for any of the options of point 2, we should ask ourselves: is the data sequential, or rather, should we take the sequence in account? ...