Anime Recommendation System Project

Deployed at https://animania-pwa.herokuapp.com

Resources

Jikan (Unofficial MAL API), documentation can be found here: https://jikan.docs.apiary.io/#
Dataset, which will be used for training the model: https://www.kaggle.com/azathoth42/myanimelist

Algorithm: Matching Users to Recommendations

User-Based Collaborative Filtering for a Personalized Recommendations List

We want to find a set of similar users and based on what similar users have rated highly or favourited, these will be matched to the user. Similarity will be defined based on cosine similarity.

Matching to General Trends

• Display a list of anime based on overall ranking or popularity (this can be all fetched through the API)
• Additional refinements include filtering genres and type of anime (i.e., TV series or movie)

User Profiles

• Email as username
• Can sign in using Google or Facebook (using Firebase authentication)
• Has a "Completed" List and a "To Watch List"
• Completed animes must be rated by a user so that recommendations can be provided
• Settings info

Main Features

• See the general trends and a personalized recommendation list
• Add watched or to watch anime into "completed" and "to watch" list respectively
• Completed animes, each having a rating (1-10), will be used to train the dataset and predict similar users
• Modify the number of recommendations given in the settings page

Tech Stack

• Languages: Python, JavaScript, HTML, CSS
• Frameworks: ReactJS (Front-End), Python Flask (Back-End)
• Database: MongoDB
• APIs/Other: JikanAPI, Google Sheets API, Firebase Authentication, Progressive Web App (PWA)