DS-GA-3001-Advanced-Python

Github repo: https://github.com/nimiw0821/DS-GA-3001-Advanced-Python

Dataset: https://grouplens.org/datasets/movielens/latest/

Small one: 100,000 ratings and 3,600 tag applications applied to 9,000 movies by 600 users.

Content-based Recommender

Time for creating a similarity matrix and generating 10 recommendations for each of the 3 movies:

Method	w/ Function Calls (s)	w/o Function Calls (s)
linear_kernel	0.896	0.858
cosine_similarity	0.937	0.905
pairwise_distances	1.473	1.406
matrix multiplication (np.dot)	0.553	0.505
matrix multiplication (operator)	0.539	0.495
nested for loops	3196.773	3098.554
nested for loops (Cython)	1478.108	1276.212

Memory-based Recommender

Method	Time in Seconds
Base	0.031
Simplified code	0.021
Numba.jit()	0.294
Cython	0.019
Combined Cython and simplified code	0.021

ALS Recommender

Time for training ALS models:

Method	Time in Seconds
Base ALS	470
ALS with Conjugate Gradient (CG)	74
ALS with CG using Cython	84
ALS with CG using Numba JIT	10