Abstract
This analysis explores the temporal sequencing of objects in a movie trailer.Temporal sequencing of objects in a movie trailer (e.g., a long shot of anobject vs intermittent short shots) can convey information about the type ofmovie, plot of the movie, role of the main characters, and the filmmakerscinematographic choices. When combined with historical customer data,sequencing analysis can be used to improve predictions of customer behavior.E.g., a customer buys tickets to a new movie and maybe the customer has seenmovies in the past that contained similar sequences. To explore objectsequencing in movie trailers, we propose a video convolutional network tocapture actions and scenes that are predictive of customers' preferences. Themodel learns the specific nature of sequences for different types of objects(e.g., cars vs faces), and the role of sequences in predicting customer futurebehavior. We show how such a temporal-aware model outperforms simple featurepooling methods proposed in our previous works and, importantly, demonstratethe additional model explain-ability allowed by such a model.