Similar Movie Finder

A program that can identify for every movie in the list, n other similar / related movies in the same list in descending order of "similarity".

Compiling and Running Project

Windows Environment :-
Compile : javac -cp .;./gson/gson-2.8.0.jar test/Test.java -d build/
Run : java -cp ./build/;./gson/gson-2.8.0.jar test.Test
Unix :-
Compile : javac -cp .:./gson/gson-2.8.0.jar test/Test.java -d build/
Run : java -cp ./build/:./gson/gson-2.8.0.jar test.Test

Input and Output

Input : "input.json" file in input directory

Output : "output.json" file in output directory

Each movie in input is of form:

  { 
    "movie_id": 1, 
    "rating": "9.3", 
    "votes": "1,318,626", 
    "description": "Two imprisoned men bond over a number ....", 
    "title": "The Shawshank Redemption", 
    "poster": "", 
    "release_date": "14 October 1994", 
    "metascore": "80", 
    "director": "Frank Darabont", 
    "storyline": "Andy Dufresne is a young and successful ...",
    "stars": [ "Tim Robbins", "Morgan Freeman", "Bob Gunton" ], 
    "year": "1994", 
    "genre": [ "Crime", "Drama" ], 
    "gallery": [ "unknown1394846836._CB379391227_.png", ], 
    "running_time": "142min"
  }

Calculation of Similarity of Movie

Let the similarity between two movies be defined using a parameter – weight

More weight -> More similarity

Then for each pair of movies, similarity factor is calculated using :
    i)   Genre              [weight = 100 for each matching genre]
    ii)  Director of movie  [weight = 50]
    iii) Rating             [weight = 10]
    iv)  MetaScore          [weight = 10]
    v)   Year of Release    [weight = 10]

Finally, the last "n" movies of the sorted list corresponding to a movie is chosen for output.

Reason behind choosing the above mentioned factors

Genre: Most important factor for similarity of two movies. Movies of same genre are more likely to be similar.
Director: A director might produce similar movie. Since the same person would be directing two movies, the movies are likely to have similar touch.
Rating: Good movies are similar to good movies and poor will be similar to poor movies. But this cannot be generalized more often, so the weight of rating is just 10% of genre.
MetaScore: Similar to audience's rating, Critics score can also be taken into account.
Year of Release: It is observed quite often that movies released in same decade are more similar. e.g., 90's of Bollywood is known for romantic and fight oriented movies. This factor is not taken into account if release date of movies differ by more than 10 years.

Procedure for calculating similarity of two movies

Let there be an imaginary link between two movies say A and B and the weight of the link represents the similarity of movies.

Increase in Weight
100 * number of similar genres
50 if director is same
10 - difference in ratings
(100 - difference in metascore)/10
10 - (difference in year of release) if both movies are released in same decade

Procedure for finding the list of similar movies

Similarity as described above is calculated for all pair of movies (250*250 in our example) and stored in a 2D array. Then the array is sorted according to similarity and the required number of similar movies is picked from last (10 in our example) for each(250) movie.