Predicting movie success using machine learning techniques

Detta är en Kandidat-uppsats från KTH/Skolan för datavetenskap och kommunikation (CSC); KTH/Skolan för datavetenskap och kommunikation (CSC)

Författare: Carl Jernbäcker; Shahrivar Pojan; [2017]

Nyckelord: ;

Sammanfattning: The area of creating predictive models using machine learning has increased in size in recent years. The market for movies is still big with hundreds of new movies created every year. The purpose of this report is to investigate whether it is possible to classify movie rating and box office revenue with metadata available before release. This was done by building a classification model with metadata obtained from the internet such as, budget and what actors are involved, etc. This study managed to correctly predict what rating a movie would have about 82% of the time using the technique with the highest success rate. The times as a model failed to predict the correct rating, it was usually by one rating group, corresponding to a deviation of approximately 17%. When a prediction of gross sales was made, it gave a positive result of 15% of the time. The results of this report are to a certain extent consistent with previous studies with similar focus in the prediction of the grade. The precision of the predictions can further be increased with a larger data set with more features. 

  HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)