Item-Item Collaborative Filtering vs Market Basket Analysis What is the basic difference between Item based Collaborative Filtering and Market Based Analysis? Is the latter a specialised case of the former? 
 A: An excellent question! One trivial difference that I can think of, is that market basket (MB) analysis considers each basket separately. So if you buy the same stuff together once a month, each time it constitutes a different basket, and it likely also contains different items each time.
However collaborative filtering (CF) considers baskets aggregated per user. So no matter how many times you buy beer and diapers together, it still counts as one vote for beer and one vote for diapers.
The other differences are more technical, such as what it is that you measure for each. In MB you care about support and confidence values and in CF, you care about a similarity measure such as cosine similarity. This is a symmetric measure. The similarity between beer and diaper is the same as the similarity between diaper and beer, but that is not the case for support/confidence.
On a conceptual level, it is possible for CF to come up with more indirect similarities such as if you buy item 1, and it finds that item 2 is bought along with it, and also that item 3 and 4 are similar to item 2. Then it can recommend them even if they are not bought along with item 1, but also with item 2.
A: @Antimony gave a perfect answer. Just wanted to add some theory that helped me to understand the difference between Item-Item Collaborative Filtering and Market Basket Analysis; as well as the applications for these two methods.
The family of algorithms used for performing market basket analysis is called association rules. Market basket analysis (or association rules) and collaborative filtering answer fundamentally different questions. Collaborative filtering can answer a question “What are the items that users with interests similar to yours like?” (Fig. 1), whereas association rules answer a question “What are the items that frequently appear together?” The answer to the first question can be used to recommend you products, videos, restaurants, hotels or any other content that you haven’t seen previously and that have been appreciated by a group of other users with interests similar to yours. The similarity of interests can be estimated from explicit indicators, for example, you and a group of other users gave same ratings to same products, or implicit indicators, for example, you and they purchased same products. Collaborative filtering is widely used for building recommender systems. However, collaborative filtering is most effective when there is a rich history of user preferences or behavior.
In the meantime, association rules can recommend you products that you will very likely purchase based on a set of products that are currently in your basket (Fig. 2). For example, if you buy a burger and fries, you will probably want soda; or a very famous example, those who buy diapers tend also to buy beer. Association rules are independent of personal preference profiles and for mining them you need a dataset of transactions from all users. Association rules and market basket analysis are generally used as an exploratory tool to mine a limited number of most common rules that can then be analysed by a human. However, association rules can also be used for building recommender systems.

Fig. 1. The illustration of collaborative filtering. Source - Wikipedia

Fig 2. A simple illustration of association rules.
