Netflix updates the suggestion algorithms taking advantage of the global growth

During the use of the Netflix service used by the suggestion algorithms they may seem simple but in reality there is a specific science behind them, based on various characteristics. But today, with the company operating in 130 countries around the world, the code behind the suggestions has seen a remarkable revolution. They have spoken Yves Raimond and Justin Basilico last technical post on the official blog, explaining the difficulties in delivering specific content to users in different countries.

The new algorithm does not use more regional models, but analyzes the types of videos you like to users around the world relying on subscribers who share the same tastes. The data is then synchronized no longer considering the state in which they live Netflix members, but suggesting what is popular at a given time among fans of a certain category. In this way, the company claims that the tips can have a level of detail never before seen on Netflix.

The new algorithm is based on four elements: availability, culture, language and monitoring. When Netflix was available only in a few regions, the suggestions were based on specific country on the assumption that in the same geographical area catalog of works was the same. If Netflix had used the old algorithm to cross the data of the various nations, then, the suggestions made would not have been certainly very helpful for the viewer.

To solve the problem, the company has updated the algorithms to recognize that subscribers can have "access to different catalogs based on geographic data and time." In practice, the tip will appear only if the members share the same item in the catalog and the same habits with the service. The most obvious advantage is that the algorithm can now use the data for a number of users extremely higher than in the past, relying on the base of international customers.

The company will continue to consider local tastes in those countries where the service is more popular and widespread, and promised that will combine the personal tastes to those premises in the future. However, the tastes of horror fans, for example, will be very similar in the various countries, and it is in these cases that the suggestions of the new automated system will be more reliable. The software then learns what are the trends over time, incrociandone data with location, language and popularity.

To ensure the effectiveness of the tips in the course of time the new system is based on a specific monitoring technique able to understand how a work is accepted in a very precise locations. "But our journey is just beginning," admitted the two Netflix engineers. "And we will try to constantly improve ourselves to make our service better."