Web version Photo by Felix Mooneeram on Unsplash
This is a movie recommender system that is based on real user ratings on movies data.
I used an unsupervised learning algorithm, the Non_negative Matrix Factorization (NMF)
The recommendation accuracy is of course subjective (it depends on each person taste) but a future improvement will be to change the NaN imputation method, the current version uses k-Nearest Neighbor imputation with 10 neighbors and weights="distance"
Here you can see a web version running in pythonanywhere using Flask.
The recommender takes three favorite movies from the user as input and outputs as many recommendations as the user wish.