The question that comes to mind when you decide to direct a movie or series is, in many cases, what am I seeing now? And the fact is that this question can be so difficult to solve that sometimes we spend too much time choosing (if we manage to solve it).
That’s why everyone today needs an intelligent streaming platform that can understand their preferences and tastes, and not just run on autopilot. From Netflix to Amazon Prime, recommender systems are becoming increasingly important as they interact directly (usually behind the scenes) with users on a daily basis.
Focusing on Netflix as the main streaming platform in Spain, despite all the big battles it is currently fighting, the platform shows very accurate recommendations, even changing the illustrations of the series depending on the user. But how does this system work?
Netflix recommendation systems cover various algorithmic approaches such as reinforcement learning, neural networks, causal models, probabilistic graphical models, matrix factorization… and many more.
And the fact is that these systems were developed by hundreds of engineers who analyze the habits of millions of users based on many factors. Every time a user accesses Netflix services, the recommendation system estimates the likelihood that the user will see a certain title based on several factors (watch time, interactions, most viewed genres…).
With all this, These are some of the key points that the platform’s recommendation system takes into account.
1. Your browsing history: This seems to be the easiest way to develop personal recommendations. And it’s that the ones you see on the platform are collected by it and then they make a list of specific recommendations.
I don’t miss you Netflix always watches what you watch. You can even upload your browsing history to it to check what you have watched in the past and when.
2. Your qualifications: Many do not perceive this factor as something important, and we are ready to bet anything that many do not even use it. Well, this is something really useful and not only for the platform, but also for you. So, you appreciate a series or a movie, Netflix will offer you one or the other.
Taking only what you’ve seen as a benchmark can lead to error, as you may not like the movie you watched over the weekend, but you only watched it for fun. Give it a bad rating and Netflix will do the rest.
3. Favorite games of other users with the same tastes: This item is pretty simple. Many times your menu will show up with people who have seen a particular series or movie that suits your tastes and has also been praised by people like you.
Xavier Amatrian, CTO of Netflix, states that the data collected “are used in various algorithms, each optimized for different purposes. Broadly speaking, most of our algorithms are based on the assumption that similar browsing patterns reflect similar user tastes.”
4. Title information: you can also see the name on your profile based on their information. And the fact is that Netflix is very smart and takes into account such details of the title as the year of release, actors, category, genre and others.
“Looking at the metadata, you can find a lot of similarities between the shows. Were they created around the same time? Do they usually have the same ratings? You can also look at user behavior when browsing, playing, searching.”Amatriain explained.
5. Time of day, devices you use and location: Finally, we highlight these very important aspects when recommending. It may seem strange, but Netflix takes all this into account, because there are differences in behavior when viewed at different times of the day and on different devices.
Location also affects recommendations. If others rate the film very highly in your area, you may also see the same in your suggestions.
All this is born when you create your account. For each new user, Netflix asks you to select the titles you would like to watch. These headings are used as a first step for subsequent recommendations. As long as the audience keeps watching Recommendations are based on the most recently viewed titles as well as the factors listed above.
Let’s add one more point: it seems that Mohammad Sabah, Netflix’s chief data officer until 2012, has given up on the idea that the platform also used your web searches. However, this is of course not 100% known to be true.
Source: Computer Hoy
