nisharma85/luther_draft

predicting the success of movies based off of tv shows

Luther

predicting the success of movies based off of TV shows

For this project is scraped box office mojo, The Numbers and wikipedia to get information about movies and TV shows. I used wolfram alpha to adjust movie budgets and revenue for inflation. After adjusting dollar values for inflation I built a linear model to predict revenue based off of budget and other features. After applying the model to all films to see which films over or underperformed there projections. Then I looked at all movies based off of TV shows to see if any if the popularity(determined buy the number of seasons or episodes that the show ran for) of the TV show or its relevance (measured by how long since the show was first on TV) could explain any of the difference between the the projected and actual success of the movie. I turns out that information about the TV shows did not explain anything. Although I did discover that just the fact that a movie was based off of a TV meant that it behaved differently than a regular movie. The difference between the regular model and the model trained only on movies based on TV shows was statistically significant. The budget of movies based on TV shows appears to have less of an effect on the revenue than it does for other movies.

The inflation notebook, creates a dictionary for converting historic dollar values into 2015 dollars. The Luther_scrape notebook scrapes the relevant information from the web and pickles it. The model_fitting notebook fits a linear model to the predict the gross of a movie in 2015 dollars. The feature selection notebook looks to see if any of the features for movies based off of TV shows can explain the over or under performance of a movie.