Before each movie’s premiere, cinema companies attempt to predict their performance in order to better allocate resources. In this study, we attempt to use deep neural networks to make these predictions on movies in Danish cinemas, both before and after the premiere. To model the movies we use static attributes, measurements of the public interest through Google Trends as well as some environmental factors. These are all drawn from publicly available sources. We compare our results both to a previous study and to those of a human expert.
Although our model makes decent predictions, it is not able to reach the same performance as the expert and has likely been affected by the narrow source of the dataset. We also examine the effect of the changing availability of Google Trends data in the days leading up to the premiere and profile which attributes had a high impact on the model’s predictions.
This project was developed for my Bachelor at the IT University of Copenhagen in a span of twelve weeks. The project was developed by a team of twoindividuals. The project was developed in Python.
Due to NDA download is not available.