Recommender System
- Python 3.7.2
- numpy
- joblib
- scipy
- six
- cython
- MF and KNN algorithm
- similarity.pyx (in recommender.py): this file has methods that are used for user's similarity measure. As a result of test, I check pearson_baseline measure is the best. pearson_baseline make use of shrunk Pearson correlation coefficient between all of users. So, I do not cosine, msd, pearson function.
- optimizing_baseline.pyx (in recommender.py): this file defines baseline_sgd function. the function computes base user and base item value by using given iteration count(n_epochs) and regularization and learning rate.
- class AlgoBase (in algo_base.py): this class is base class for MF class and KNN class.
- class Prediction (in algo_base.py): this class is used as a prediction instance that contains uid, iid, real rating, estimated rating information.
- RMSE