Classic recommendation algorithms implementation
| DNN Model | RMSE | MAE |
|---|---|---|
| NeuMF | 0.9433 | 0.7485 |
| MF Model | RMSE | MAE |
|---|---|---|
| Baseline | 0.946 | 0.742 |
| SVD | 0.931 | 0.731 |
| SVDPlusPlus | 0.927 | 0.726 |
| Explicit ALS | 1.199 | 0.903 |
| Implicit ALS | 2.752 | 2.525 |
| Neighborhood Model | RMSE | MAE |
|---|---|---|
| Itemcf | 1.029 | 0.802 |
| WeightedSlopOne | 1.043 | 0.835 |
import os
from util.databuilder import DataBuilder
from algorithm.dnn.neumf import NeuMF
file_name = os.path.abspath("data/ml-100k/u.data")
data_builder = DataBuilder(file_name, just_test_one=True)
data_builder.eval(NeuMF(epochs=2), k_folds=5)- MovieLens
- Neural Collaborative Filtering
- Yehuda Koren. Factorization meets the neighborhood: a multifaceted collaborative filtering model
- Matrix factorization techniques for recommender systems
- Advances in Collaborative Filtering
- Slope one predictors for online rating-based collaborative filtering