An overview of a recommendation engine

We will now focus on situations where users have provided rankings or ratings on previously viewed or purchased items. There are two primary categories of designing recommendation systems: collaborative filtering and content-based (Ansari, Essegaier, and Kohli, 2000). The former category is what we will concentrate on as this is the focus of the recommenderlab R package that we will be using.

For content-based approaches, the concept is to link user preferences with item attributes. These attributes may be things such as the genre, cast, and storyline for a movie or TV show recommendation. As such, recommendations are based entirely on what the user provides as ratings; there is no linkage to what anyone else recommends. This has the advantage over content-based approaches: when a new item is added, it can be recommended to a user if it matches their profile instead of relying on other users to rate it first (the so-called first rater problem). However...