Recommender systems in Machine Learning

What are recommendation engines?

It is the most powerful and useful application of machine learning technology in business. these days. Now a day every big giant’s company like Google, Amazon, Netflix uses recommendation engine.

What are recommendation engines?

Recommender system most relevant item to the user by filtering the data using different algorithms.

Note: Recommender System recommends the relevant items to the user based on items he has liked/based on the historical data about the users.

Let us take one example: if you are new to the youtube/ Netflix or you visit the e-commerce website, that time that website or that system doesn’t have any historical data about the user or then how does that website recommend the best movie/videos or products to the user. So one possible solution is to recommend the products/movies/videos based on top ratings or based on top sales if it is e-commerce.

It is similar to Netflix, Youtube like you watch the videos then there are some recommended videos that are similar to the videos that you have watched.

Note: Now a day recommender system is very useful and seems very profitable for business problems. It creates positive impacts on the user’s experience by recommending the videos or the products that the user may like.

Lets us look out how it works

The main task of building a recommender system is the given a user I if I already know that the user has liked or watched some videos/ products i.e item 1, item 2, item 3 and so on then I want to recommend the new item that most likely the user I will like.

This is the main task of the recommender system.

Machine learning algorithms in recommender systems are typically classified into two categories

1. content-based.

It is something similar to classification and regression technique. It does not use the rating information. it uses the attribute of items/users.

This system uses item metadata, such as genre, director, description, actors, etc. for movies, to make these recommendations. The general idea behind these recommender systems is that if a person liked a particular item, he or she will also like an item that is similar to it.

The point of content-based is that we have to know the content of both user and item. Usually, we construct user-profile and item-profile using the content of shared attribute space. For example, for a movie, you represent it with the movie stars in it and the genres (using a binary coding for example). For user profile, you can do the same thing based on the users likes some movie stars/genres etc.

That is — In this recommendation process, the engine compares the items that are already positively rated by the user with the items he didn’t rate and looks for similarities. Items similar to the positively rated ones will be recommended to the user. Here, based on the user’s taste and behavior a content-based model can be built by recommending articles relevant to the user’s taste.

The content-based recommendation engine will only recommend articles related to these categories and may never recommend anything in other categories as the user never viewed those articles before while in collaborative filtering A user gets recommendations for those items that user hasn’t rated before but was positively rated by users in his/her neighborhood.

2. collaborative filtering methods.

The collaborative algorithm uses “User Behaviour” for recommending items. They exploit the behavior of other users and items in terms of transaction history, ratings, selection, and purchase information. Other users behavior and preferences over the items are used to recommend items to the new users.

let’s take an example for collaborative filtering based method

user 1 = movie 1, movie 2, movie 3

user 2 = movie 1, movie 3, movie 4

user 3 = movie 1, ?

So as we can see that user 1 and user 2 both like movie 1 and movie 3, so suppose we want to recommend the movie for the user 3 then we can recommend movie 3 because movie 3 is liked by both user 1 and user 2 and there is a high chance that user 3 will also like movie 3.

The core idea is the user who agreed in the past tends to also agreed in the future and this is the core idea/assumption of collaborative filtering.

Because as we can see that user 1 in the past liked the movie 1 and also user 1 and user 2 in the past liked the movie 1 so user 1,2 and 3 agreed that movie 1 is a great movie. so in the future, there is a high chance that user 3 will like movie 3 which is user 1 and user 2 agreed that movie 3 is a great movie. so this is the idea/ assumption made my Collaborative filtering.

Thanks for reading