Repository for the movie_recommendation project in OC Data Scientist path
Main objective : Create a movie recommendation tool using unsupervised exploration
This project has been tackled in two separate phases :
- Phase 1) Early project in the DS path, with a 5000 films database (IMDb) we aim to develop a recommender tool using unsupervised learning from the dataset (nearest neighbors algorithm)
- Phase 2) Last project of the DS path, with 53000 films of this database (IMDb) and only using genres and keywords, we train a neural network with two embeddings layers using Will Koehrsen's work (https://towardsdatascience.com/neural-network-embeddings-explained-4d028e6f0526) on book recommendation for our own movie's recommendation purpose. Results are better than I was expected and the recommender, through the powerful non-linear behaviour learning of the neural network, is outputting some accurate-enough recommendations