Yak shaving a good place to eat using non negative matrix factorization
summary
Loads a user‑by‑restaurant rating matrix and applies non negative matrix factorization in Python, using 11,000 users × 500 restaurants and settling on 10 latent factors. The code produces restaurant and user factor weights, maps factors to interpretable signals (pinchos, toast, pizza styles), and finds similar venues and like‑minded users. It is a proof‑of‑concept that NMF suggests plausible matches, but accuracy depends on sparsity, factor count and tuning.
AI-written summary · about
chapters (4)
related talks
39mDiscover your latent food graph with this 1...
35mData Science for Dinner: Recommending Recipes
01h 31mModern NLP in Python
20mJeremy Schiff, Senior Manager, Data Science, OpenTable
40mMachine Learning over Small Data: generating new variables with predictive power
27mOur experiments with food recommendations
38mRecommender systems with Tensorflow
32mLeveraging recommender systems to personalise search results