Machine Learning is Fun – Part 15 – Recommendation System

Viewer Rating
Hello! I hope you enjoyed my last few articles on Machine learning. Today, I want to talk about a very interesting topic, that is recommendation system. There are a number of articles on this topic. But I do not want to re-explain any of these. My intention is to make you understand the techniques that you need to consider before building a recommendation system.
  • You are an employee of a movie or article based company where they want you to build a recommendation system.
  • You are working in a retail company where they want you to build a product recommendation system.
  • You are working in a hotel or tourism company where they want you to build a hotel recommendation system.
  • Well, what would be the first step?
  • Collect data of users and related products. But what kind of data?
  • Understand the business requirement and user perspective.
  • Case 1: Suppose you are an employee of Netflix. Now, you have collected user / movie data like what the list of movies which are watched mostly by which users, what kind of rating has been given by users, feedback etc. The 2nd set of data which is company’s static data like movie genre: action, comedy, adult, kids, romantic etc. The 3rd set of data: demographic content of users. Well, what kind of recommendation system should you use? Content based or collaborative filter based? Or maybe, hybrid? Well, for movie selection purpose, users’ demographic content always plays a huge role right? Like, for example, 10-15 year old kids always watch mostly kids-content based movies while men in the 25-35 age group mostly like to watch action-genre movies, women aged 15-30 years prefer romantic movies etc. So, we can go for user content based recommendation system. Next, we have fix set of genres for any kind of movie – right? So we can also go for movie genre content based recommendation system. Now, we also have a trend of giving ratings on the movie that we have watched. So we can go for collaborative rating-based recommendation system. I hope you liked the discussion. I do not want to explain how you should implement these recommendation systems because you can easily get that information on the internet. But before jumping to build any recommendation system, you must first plan your design. I shall discuss on the case 2 design in my next article. Stay tuned.

    2 thoughts on “Machine Learning is Fun – Part 15 – Recommendation System”

    Leave a Comment

    %d bloggers like this: