Anime-Recommendation

Anime Recommendation System using data collected from Jikan's unofficial MyAnimeList API v3. Used sklearn Python library to apply k-nearest-neighbour algorithm to train a recommendation model for all titles (up to 26th of July, 2021) on myanimelist.com. Type in any title on MyAnimeList to get 10 closest recommendations.

Instructions to use

If all you need is the recommendation system, then only src/recommendation.py, data/anime.csv, and data/users.csv is needed to run.

get_users_from_clubs.py

• Collects a list of usernames from specified clubs to a .txt file
• usage: get_users_from_clubs.py [-h] -o OUTPUT
• Example: get_users_from_clubs.py -o usernames.txt

get_user_ratings.py

• Collects user ratings from list of usernames generated from get_users_from_clubs.py
• Outputs numbered and segmented .json files starting with index of FILE_COUNT, each with a specified number of users
• usage: get_user_ratings.py [-h] -o OUTPUT -i INPUT -c FILE_COUNT -n NUMBER_OF_USERS_PER_FILE
• Example: get_user_ratings.py -o users.json -i usernames.txt -c 1 -n 500

collect_all_anime.py

• Collects all anime and their relevant information from a specified range of anime ID
• Outputs numbered and segmented .json files starting with index of FILE_COUNT, each with a specified number of titles
• usage: collect_all_anime.py [-h] -o OUTPUT -s START_ID -e END_ID -c FILE_COUNT -n NUMBER_OF_ANIME_PER_FILE
• Example: collect_all_anime.py -o anime.json -s 1 -e 54000 -c 1 -n 100

combine_json.py

• Outputs a combined .json file from the numbered and segmented .json files generated from get_users_from_clubs.py and collect_all_anime.py
• Specify the file format of the segmented .json files and specify the first and last index of the files to be combined
• usage: combine_json.py [-h] -i INPUT_FILE_FORMAT -o OUTPUT -s START_INDEX -e END_INDEX
• Example: combine_json.py -i users.json -o combined_users.json -s 1 -e 30

users_json_to_csv.py

• Converts combined_users.json generated from combine_json.py to suitable .csv format for recommendation.py to use
• usage: users_json_to_csv.py [-h] -o OUTPUT -i INPUT
• Example: users_json_to_csv.py-o combined_users.csv -i combined_users.json

anime_json_to_csv.py

• Converts combined_users.json generated from combine_json.py to suitable .csv format for recommendation.py to use
• usage: anime_json_to_csv.py [-h] -o OUTPUT -i INPUT
• Example: anime_json_to_csv.py -o combined_anime.csv -i combined_anime.json

recommendation.py

• Gives anime recommendations
• If no input .pkl model is inputted, you will have to specify the output path to the generated model.
• usage: recommendation.py [-h] (-o OUTPUT_MODEL | -i INPUT_MODEL) -a ANIME_CSV -u USERS_CSV
• Example: recommendation.py -o knn_model.pkl -a combined_anime.csv -u combined_users.csv

To do

• Add option to filter by genre, age rating, type (eg. TV, Movie,...), popularity
• Collect more users for more accurate recommendations (currently ~ 15k users collected)