"""
Baseline only
"""
from surprise import BaselineOnly
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
param_grid = {'bsl_options': [{'method': 'als', 'n_epochs': 30},
{'method': 'sgd', 'learning_rate': 0.0007}]}
recommender = Recommender(algorithm=BaselineOnly,
param_grid=param_grid,
dump_model=True,
dump_file_name=get_dump_path('baseline_only'))
recommender.recommend(uids=uids, verbose=True)
"""
Slope One
"""
from surprise import SlopeOne
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
recommender = Recommender(algorithm=SlopeOne,
dump_model=True,
dump_file_name=get_dump_path('slope_one'))
recommender.recommend(uids=uids, verbose=True)
"""
Neighborhood-based collaborative filtering (kNN-with-means)
"""
from surprise import KNNWithMeans
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
param_grid = {
'k': [20, 40],
'sim_options': [{
'name': 'msd'
}, {
'name': 'cosine'
}, {
'name': 'pearson'
}, {
'name': 'pearson_baseline'
}, {
'name': 'pearson_baseline',
'shrinkage': 150
}]
}
recommender = Recommender(algorithm=KNNWithMeans,
param_grid=param_grid,
dump_model=True,
dump_file_name=get_dump_path('knn_with_means'))
recommender.recommend(uids=uids, verbose=True)
from surprise import SVD
from rs import Recommender, load_data_from_file, parse_config, get_dump_path
data_path = parse_config(section='Path', key='data')
votes = load_data_from_file(data_path + '/votes.csv', (-1, 1))
clips = load_data_from_file(data_path + '/clips.csv', (0, 1))
print('■ Voting data')
param_grid = {'n_epochs': [20, 30], 'n_factors': [20, 50]}
recommender = Recommender(algorithm=SVD,
param_grid=param_grid,
data=votes,
rating_threshold=0.5,
dump_model=True,
dump_file_name=get_dump_path('viblo_votes'))
recommender.recommend(uids=[2, 9, 21], verbose=True)
print()
print('■ Clipping data')
param_grid = {'n_epochs': [20, 30], 'n_factors': [20, 50]}
recommender = Recommender(algorithm=SVD,
param_grid=param_grid,
data=clips,
rating_threshold=1,
dump_model=True,
dump_file_name=get_dump_path('viblo_clips'))
recommender.recommend(uids=[2, 12825, 13072], verbose=True)
"""
Matrix factorization - Non-negative Matrix Factorization
"""
from surprise import NMF
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
param_grid = {'n_epochs': [50, 100], 'n_factors': [15, 20]}
recommender = Recommender(algorithm=NMF,
param_grid=param_grid,
dump_model=True,
dump_file_name=get_dump_path('nmf'))
recommender.recommend(uids=uids, verbose=True)
import json
from rs import Recommender, load_data_from_file, parse_config
path = parse_config(section='Path', key='data') + '/views_without_ips.csv'
views = load_data_from_file(path, (1, 238))
recommender = Recommender(data=views, rating_threshold=10, anti_testset=False)
data = recommender.recommend(uids=[1087], n_items=15, verbose=True)
with open('data.json', 'w') as file:
json.dump(data, file)
"""
Neighborhood-based collaborative filtering (kNN-basic)
"""
from surprise import KNNBasic
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
param_grid = {'k': [20, 40],
'sim_options': [{'name': 'msd'},
{'name': 'cosine'},
{'name': 'pearson'},
{'name': 'pearson_baseline'},
{'name': 'pearson_baseline', 'shrinkage': 150}]}
recommender = Recommender(algorithm=KNNBasic,
param_grid=param_grid,
dump_model=True,
dump_file_name=get_dump_path('knn_basic'))
recommender.recommend(uids=uids, verbose=True)
"""
Co-Clustering
"""
from surprise import CoClustering
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
param_grid = {'n_epochs': [20, 60], 'n_cltr_u': [3, 5], 'n_cltr_i': [3, 5]}
recommender = Recommender(algorithm=CoClustering,
param_grid=param_grid,
dump_model=True,
dump_file_name=get_dump_path('co_clustering'))
recommender.recommend(uids=uids, verbose=True)
"""
Matrix factorization - SVD++ using Alternating Least Squares
"""
from surprise import SVDpp
from rs import Recommender, get_dump_path
uids = [1, 2, 3]
param_grid = {'n_epochs': [10, 20], 'reg_all': [0.01, 0.02]}
recommender = Recommender(algorithm=SVDpp,
param_grid=param_grid,
dump_model=True,
n_folds=3,
dump_file_name=get_dump_path('svdpp'))
recommender.recommend(uids=uids, verbose=True)