The developments of today’s technologies, make recommendation systems become popular. Many e-commerce companies can offer products to their customers that they may like by using their recommendation systems.

What are the recommendation systems?

The recommendation systems are based on information filtering systems that aim to predict the amount of preference a user gives to an item.
Thanks to these systems, we can offer meaningful products to the user by examining the data we have on a user basis and by extracting useful information. Some examples such as showing similar products when we shop on Amazon and presenting similar products when we watch TV shows on Netflix can be given.

There are different types of recommendation systems.

1-Collaborative Filtering System

Estimates are made based on the choices of users or products that are similar to each other. Assume that the first user has watched and liked the TV series “Friends”, “Modern Family” and “Breaking Bad”, while the second user watched and liked the dramas “Friends” and “Modern Family”. When we looked into the series that the users have watched and liked, we can conclude that their tastes are similar and the second user will also like the ‘Breaking Bad’ series.

We can examine the subcategories of Collaborative Filtering Systems in two groups as User-Based and Item-Based.

User-Based: Recommendations are made based on the similarity of users. The user-item matrix is ​​created.

Item-Based: Recommendations are made based on selected products. An item-by-item matrix is ​​created.

2-Content-Based Filtering System

Recommendations are developed based on the similarities of the product ingredients.
For example, the user watched a movie about animals. We can suggest movies about animals for the next movie the user will watch.

3-Popularity Based Recommendation System

Generally, preferred products are offered instead of user-based products. The best seller products on the e-commerce site or the most-watched movie on the movie site will be recommended to users.

4-Hybrid Recommendation System

This system is being applied by considering both content and choices of similar users. Additionally, it may require more data than others. But more favorable results can be obtained.