Skip to content

Latest commit

 

History

History
82 lines (55 loc) · 2.75 KB

README.md

File metadata and controls

82 lines (55 loc) · 2.75 KB
Raw

Anime Recommender

Web App: https://anime-recommender-AlimU.pythonanywhere.com (not fully mobile optimized)

Parametrized anime recommendation system that produces recommendations based on

  1. individual title or set of titles img.png
  2. user profile (refer to Current limitations section for further details) img_1.png

Local Setup

  1. Clone the repository
git clone https://github.com/AlimU11/Anime-Recommender.git
  1. Install requirements from requirements.txt
pip install -r requirements.txt
  1. Run main_etl.py
python main_etl.py

This will fetch the data from the source and save it in the data folder. This should take approximately 15 minutes. See

python main_etl.py -h

to run only chosen parts of the ETL process.

  1. Run main_local.py
python main_local.py

This will start the local server with the app.

Docker Setup

You can also run the app in a docker container. Check out this page for more information.

File Structure

  1. anime_recommender module - contains all the code for the app

  2. assets - images used in README.md and NOTES.md. CSS and js for dash app

  3. notebooks - notebooks used for data exploration and model development

  4. data - folder with data used for the app. Not included in the repository due to its size. Could be easily recreated by running `main_etl.py

Roadmap

Key features in further updates (unordered, full details are at the project page):

  1. Add new recommendation engines
  2. Improve UI - add dark theme
  3. Make responsive layout for mobile devices Done ✔️ assets/img_3.png
  4. Create weekly or monthly etl process to update data mart Done ✔️
  5. Resolve current limitations (refer to Current limitations section for further details)
  6. Add support for user profiles from other sites (MyAnimeList, Kitsu, etc.)
  7. Create etl and data mart for manga

Current limitations

  1. User's lists - all titles in INCLUDED user's lists MUST HAVE scores
  2. Score systems - other than ten-point score system was not yet tested
  3. Only top 20 results are shown
  4. Only titles with status 'Completed' at the moment of data scrapping (20.09 - 21.09) are taken into account and could be recommended
    Only titles with status 'Completed' at the moment of data update (every week on Monday at 04:00 UTC) are taken into account and could be recommended
  5. Only AniList user profiles supported
  6. Further limitations might be mentioned at the project page