Artificial intelligence has learned to predict the popularity of movies

Искусственный интеллект научился прогнозировать популярность фильмов

Neural network prediction of the possible success of the film and its likely popularity, developed kinokompanija 20th Century Fox.

The algorithm is required to analyze not even the whole movie, and its trailer. The visual objects and the time sequence of the objects in the trailer of the film can transmit information about the type of film, the plot, characters and directing.

In combination with historical data analysis can make predictions of the behavior of the audience. For example, the viewer can buy tickets for the new film, if he’s seen movies with a similar plot in the past.

Understanding the audience composition is important for the studios that invest in the creation of films with uncertain commercial outcome. One of the solutions of uncertainty might be that the Studio will now know how it will look in the film is commercially a few months before its release. Another related source of uncertainty is the audience.

An important feature in studios is the understanding with micro-segmentation of the customer base. For example, not all superhero films have the same audience, etc. In recent years, the Studio has invested in the tools for learning and

display customer segments and creating predictions for future movies.

Granular forecasts at the customer level are commonplace in important business decisions and provide a reliable barometer of potential financial performance of the film.

Recommendation system for the production of films have become valuable tools that are particularly well suited to provide long-term forecasts to specific audiences to support decision-making in relation to the launch and positioning of films.

MERLIN, recommendation system for theatrical releases, built by 20th Century Fox, is used to predict attendance of spectators and index segment for the year ahead, as well as to Refine the forecast with signals of user behavior.

Prediction of user behavior much earlier than the release of the film, is an example of pure prediction and is difficult

for new films that are not sequels, and movies that transcend traditional genres.

The results showed that the system of recommendations based on the analysis of object sequences are quite promising.

Source: VistaNews

Share Button