The model used is a collaborative filtering based recommender system built using Tensorflow 2.0 on this Kaggle dataset.
This WebApp primarily uses Express,React and Node.
A version is deployed here. Currently the backend is hosted on Glitch which takes a lot of time to wake up and process (~2 minutes).
View a show's details and get recommendations for similar shows to watch.
This data is obtained from the MAL dataset and AniList API.
Provide their ratings for some shows and get recommendations based on those ratings. These can be entered manually or imported from MAL.
Jikan is used to get user ratings.
The ratings are used to train neural network, followed by making predictions from this network. It takes a lot(~2 minutes) of time to train. So, I have used Bull to queue jobs. To track job progress, polling is used.
Search for anime using keywords and tags.
I have used FuseJS for fuzzy searching. Tags were in the MAL dataset.
- Remove duplicates from recommendations
- Speed up the neural network
- Dark theme