First, I would like to say that I understand the difference between a classification problem and a regression one.
Let's imagine that I'm trying to predict the time of a sprinter to run 100 meters based on his nationality, his height, his weight, etc.
I could use regression to predict a time as close as possible. But I could also create a class for the sprinters which run in less than 9.5 seconds, another class for the sprinters which run in less than 10 seconds, etc. I would then try to predict the class of each runner.
My question is the following: Even though it is hard to compare the results of a classification problem with the ones of a regression one, should I start by addressing the problem as a classification problem or a regression one ?