NMF #3
Description
Simple NMF
- Decompose user x anime matrix into 20 groups
- Get group row with scores
-
argsortto get likely hood of every anime already ranked
NMF + cos sim
- cosine similarity with in-group users
- rank users according to similarity and get top K
- Get top K users best ranked movies excluding not-rated
- w1 * norm(group_score) + w2 * norm(top_k_score)
- optimize [w1, w2]