추천 R패키지 속도 및 성능 비교 연구 결과입니다.
- 데이터셋 : MovieLense (100만행)
- 컴퓨터 사양 : 7i, 32Gb RAM
패키지 목록
package name |
package description |
Myrrix |
Real-Time, Scalable Clustering and Recommender System, Evolved from Apache Mahout |
recommenderlab |
Lab for Developing and Testing Recommender Algorithms |
recosystem |
Recommender System using Matrix Factorization |
rrecsys |
Environment for Evaluating Recommender Systems |
slimrec |
Sparse Linear Method to Predict Ratings and Top-N Recommendations |
패키지 성능 비교
package name |
algorithm |
time(min) |
RMSE |
recommenderlab |
Most Popular |
4.27 |
0.9725 |
|
User-Based CF |
5.03 |
1.0464 |
|
Item-based CF |
7.11 |
1.5074 |
|
SVD |
5.52 |
1.0204 |
|
Funk SVD |
13.91 |
0.9106 |
|
Random |
3.49 |
1.3832 |
|
ALS |
13.14 |
0.9032 |
rrecsys |
itemAverage |
7.37 |
0.9614 |
|
userAverage |
6.95 |
1.0140 |
|
globalAverage |
6.22 |
1.0913 |
|
IBKNN |
7.53 |
1.0853 |
|
UBKNN |
37.49 |
1.0196 |
|
FunkSVD |
31.36 |
1.0811 |
|
SlopeOne |
15.48 |
0.9028 |
recosystem |
Matrix Factorization |
0.68 |
0.8529 |
slimrec |
Sparse Linear Method |
25.52 |
2.2196 |
SVDApproximation |
SVDApproximation |
4.92 |
0.9313 |
SmartCat-labs’s Git R code |
ibcf |
1.76 |
0.8859 |
|
ubcf |
1.74 |
0.8564 |
전체 분석 코드