In order to be able to complete one of my pet projects, a program that automatically recommends movies to me, I am trying to come up with a set of requirements for a movie to (probably) be one I will like. I’m thinking of using a points system; awarding points to movies for various aspects. 100 points is the ideal goal. Negative points are possible.
To help me manage this rating system, I did what any sane person would do: I made a spreadsheet. I currently have point measures for movie length, IMDb rating, and awards earned (Grammy, Oscar, and Golden Globe only!) I also assign bonus points based on genre, director, and actor. The way I’m exporting the spreadsheet doesn’t seem to let you view the formulas used, so I listed them next to the–well, you should just look at the spreadsheet.
In case you don’t, the biggest modifier is the movie’s IMDb rating, which is calculated by (rating – 5) * 15. Thus, it actually gives negative points if the movie is rated below 5! Harsh. I find ratings to be the biggest indicators of how good a movie is, so if a movie has a good IMDb rating (8.5 or more, say) it can get upwards of 50 points from the IMDb rating alone.
It currently seems to work well for movies at the high end of the spectrum (Inglorious Basterds, The Godfather), but for movies that are out of my usual action-adventure zone (such as Up in the Air) are rated too low, because I assign negative scores to Romance and Comedy categories. Maybe I’ll just have to find out about those types of movies myself.
I also plan on incorporating a Bayes classifier to add or subtract points to my rating based on what it finds…but that’s pretty far into the future.
