Recently, I studied several clustering and prediction papers. Most of them use terms "precision" and "recall" as references of accuracy. Some of them explain what these two terms mean, but unfortunately, I feel difficult to fully interpret their explanation. Thus, I would like a more a plain description to these two terms:
assume you have a set of values called V, your algorithm is going to pick some values out of the dataset called P. In the dataset, you have W values is what you really want. In the values you pick out, Q values is what you really want. Therefore,
precision = Q/P
recall = Q/W
I feel this kind of explanation is more intuitive to me. How do you think? :p
