This project is a content-based filtering approach to video game recommendation & exploration, revolving around a core ML-Engine Flask application, and API for model deployment.
The project uses Doc2Vec and TFIDF word embeddings to map each product within the genre space
Visit jr-recommender.herokuapp.com to try it out!
Roadmap:
Development version experiments with KD Tree Nearest Neighbour search for content based recommendation of TFIDF-tag embeddings
Embed TFIDF-tags in mid-sized dimension space and use cosine similarity
A small collaborative-filtering data set was found, we will investigate this avenue!