In Russia, the first neural network has started operating, which recommends viewers to choose a certain movie or TV show to watch. The development of the Big Data center of MTS and KION includes all movies and TV shows that have ever been presented on the cinema service. The selection includes even those film works that are no longer shown in Russia.

The user can choose movies based on interests at any time, and the neural network recommends what to watch based on the analysis of current preferences.
The user can choose movies based on interests at any time, and the neural network recommends what to watch based on the analysis of current preferences.
We decided to create a new tool that will help users choose the most suitable movies and TV shows for themselves - for example, if they want to watch something new or need to choose one movie for a company of several people. The neural network Explore KION allows this, and its main value for us as a service is that it saves users time choosing a movie.
CEO of MTS Media Alexei Ivanov

Explore KION is based on what film works the user liked.

The selection can be made not only from movies but also from TV shows.
The selection can be made not only from movies but also from TV shows.

After making a choice, the neural network, based on the user's current preferences, will select a list of relevant paintings that can be watched on KION.

You can enter the title yourself, or you can see what other users choose the most.
You can enter the title yourself, or you can see what other users choose the most.

Explore KION works with two levels of the algorithm. The candidate selection model selects movies and TV shows that the user might like based on what he chose.

The genre, keywords of content, and what people often watch with a certain movie affect the recommendations.
The genre, keywords of content, and what people often watch with a certain movie affect the recommendations.

Then movies and TV shows selected on the first level and interaction history with cards from the current session go into the ranking model.

The model assesses the relevance of each movie candidate based on distances from this candidate to movies and TV shows that the user added to the selection, or didn't choose at all.
The model assesses the relevance of each movie candidate based on distances from this candidate to movies and TV shows that the user added to the selection, or didn't choose at all.

The algorithm will also work when it's necessary to choose a movie for a company of viewers with different interests.

Sources :
KION

Сейчас на главной