Learning How to Rank and Collecting User Behavior

Till now we were looking at recommendationsRecommendation as a rating prediction problem, but an interesting and alternative approach would be to see if it can now be treated as a ranking problem, i.e., how to arrange, display or stack the results in some particular order. So sometimes it may make more sense if we stack the most relevant choice at the top, followed by the second most relevant choice at the second place, and so on. In this chapter, we take a look at the ranking methods used by FourSquare and Facebook, as well as some LtR (learning to rank algorithmsAlgorithm). Many of these algorithmsAlgorithm have been first used in IR (Information Retrieval), and many of these algorithmsAlgorithm have also been used quite successfully without much hindrance. While on the topic of ranking, we also need to be careful about fakeFake user profiles who deliberately give biased feedbacksFeedback to increase or decrease the rank of an object. So here we also take a look at how to collect the likes and dislikes of a user and how to filter fakeFake profiles.