/movie_rec

Primary LanguageJupyter Notebook

Background

A movie recommendation engine using the Steamlit library for webgui. Recommendations are based off of a score calcuated as:

  1. var1 = Users who liked (rated >= 4 stars) the given movie (the movie that was searched for)
  2. var1_recs = The other movies in our database that these users also liked
  3. perc1_recs = Percent of var1 users that liked each of the recommendations, threshold of at least 10%
  4. var2 = Users who liked each of the var1_recs
  5. perc2_recs = Percent of var2 users that liked each of the recommendations, no threshold
  6. score = perc1_recs/perc2_recs, or the share that var1 users make up of all the users that liked each movie

Example

Searching for toy story will give back Wreck-It Ralph (2012) as the top recommendation, with an average user rating of 3.7 stars and a score of 5.07. The score says that 5x as many users liked Wreck-It Ralph if they also liked Toy Story 3.

A higher score should be more likely to reflect the unique likes of users who also liked the given movie