def recommender_setup_train():
recommenders = []
topPop = TopPop(URM_train)
topPop.fit()
topPopRecommender = recommenderEntry(topPop, "topPop")
recommenders.append(topPopRecommender)
elasticNet = ElasticNetRecommender(URM_train)
elasticNet.fit("train")
elasticNetRecommender = recommenderEntry(elasticNet, "elasticNet")
recommenders.append(elasticNetRecommender)
pureSVD = PureSVDRecommender(URM_train)
pureSVD.fit()
pureSVDRecommender = recommenderEntry(pureSVD, "pureSVD")
recommenders.append(pureSVDRecommender)
sslim = ReccomenderSslim(URM_train)
sslim.fit(train="-train")
sslimRecommender = recommenderEntry(sslim, "sslim")
recommenders.append(sslimRecommender)
CFItem = ItemKNNCFRecommender(URM_train)
CFItem.fit(shrink=25, topK=10, similarity="jaccard", feature_weighting="TF-IDF", normalize=False)
CFItemRecommender = recommenderEntry(CFItem, "CFItem")
recommenders.append(CFItemRecommender)
slim = SLIM_BPR_Recommender(URM_train)
slim.fit(path="../../", train="-train")
slimRecommender = recommenderEntry(slim, "slim")
recommenders.append(slimRecommender)
CFUser = UserKNNCFRecommender(URM_train)
CFUser.fit(703, 25, "asymmetric")
CFUserRecommender = recommenderEntry(CFUser, "CFUser")
recommenders.append(CFUserRecommender)
CBItem = ItemKNNCBFRecommender(URM_train, ICM_all)
CBItem.fit(shrink=120, topK=5, similarity="asymmetric", feature_weighting="none", normalize=True)
CBItemRecommender = recommenderEntry(CBItem, "CBItem")
recommenders.append(CBItemRecommender)
P3a = P3alphaRecommender(URM_train)
P3a.fit(alpha=0.25662502344934046, min_rating=0, topK=25, implicit=True, normalize_similarity=True)
P3aRecommender = recommenderEntry(P3a, "P3a")
recommenders.append(P3aRecommender)
P3b = RP3betaRecommender(URM_train)
P3b.fit(alpha=0.22218786834129392, beta=0.23468317063424235, min_rating=0, topK=25, implicit=True,
normalize_similarity=True)
P3bRecommender = recommenderEntry(P3b, "P3b")
recommenders.append(P3bRecommender)
return recommenders
def filter_seen(self, user_id, scores):
start_pos = self.test.indptr[user_id]
end_pos = self.test.indptr[user_id + 1]
user_profile = self.test.indices[start_pos:end_pos]
scores[user_profile] = -np.inf
return scores
URM_train, URM_test = train_test_holdout(URM_all, train_perc=0.8)
URM_train, URM_validation = train_test_holdout(URM_train, train_perc=0.9)
itemColl = ItemKNNCFRecommender(URM_train)
#itemColl.fit(shrink=106, topK=63, similarity="jaccard")
userColl = UserKNNCFRecommender(URM_train)
userColl.fit(shrink=100, topK=3, similarity="cosine")
itemCont = ItemKNNCBFRecommender(URM_all, ICM_all)
#itemCont.fit(shrink=120, topK=5, similarity="jaccard")
pureSVD = PureSVDRecommender(URM_all)
#pureSVD.fit()
hybridrecommender = ItemKNNScoresHybridRecommender(URM_train, userColl,
userColl, userColl,
userColl)
# rank items
ranking = scores.argsort()[::-1]
return ranking[:at]
def filter_seen(self, user_id, scores):
start_pos = self.URM.indptr[user_id]
end_pos = self.URM.indptr[user_id + 1]
user_profile = self.URM.indices[start_pos:end_pos]
scores[user_profile] = -np.inf
return scores
itemColl = ItemKNNCFRecommender(URM_all)
itemColl.fit(shrink=50, topK=10)
elasticNet = SLIMElasticNetRecommender(URM_all)
elasticNet.fit()
users = utils.get_target_users("../../Dataset/target_users.csv")
hybridrecommender = ItemKNNScoresHybridRecommender(URM_all, itemColl, elasticNet)
hybridrecommender.fit(0.3)
with open("../../../Outputs/itemColl+elasticNet_0.3.csv", 'w') as f:
f.write("user_id,item_list\n")
for user_id in users:
f.write(str(user_id) + ", " + utils.trim(hybridrecommender.recommend(user_id)[:10]) + "\n")
hybridrecommender.fit(0.5)
from External_Libraries.Base.Evaluation.Evaluator import EvaluatorHoldout
from External_Libraries.ParameterTuning.SearchBayesianSkopt import SearchBayesianSkopt
from skopt.space import Real, Integer, Categorical
from External_Libraries.ParameterTuning.SearchAbstractClass import SearchInputRecommenderArgs
URM_all = sps.csr_matrix(sps.load_npz("../../../../Dataset/old/data_all.npz"))
"""
URM_train = sps.csr_matrix(sps.load_npz("../../../../../Dataset/data_train.npz"))
URM_test = sps.csr_matrix(sps.load_npz("../../../../../Dataset/data_test.npz"))
"""
ICM_all = sps.csr_matrix(sps.load_npz("../../../../Dataset/ICM/ICM_all.npz"))
URM_train, URM_test = train_test_holdout(URM_all, train_perc=0.8)
URM_train, URM_validation = train_test_holdout(URM_train, train_perc=0.9)
itemKNNCF = ItemKNNCFRecommender(URM_train)
itemKNNCF.fit(shrink=25,
topK=10,
similarity="jaccard",
feature_weighting="TF-IDF",
normalize=False)
itemKNNCBF = ItemKNNCBFRecommender(URM_train, ICM_all)
itemKNNCBF.fit(shrink=120,
topK=5,
similarity="asymmetric",
feature_weighting="none",
normalize=True)
W_sparse_CF = itemKNNCF.W_sparse
W_sparse_CBF = itemKNNCBF.W_sparse
pyplot.plot(x_tick, MAP_per_shrinkage)
pyplot.ylabel('MAP')
pyplot.xlabel('Shrinkage')
pyplot.savefig("shrink.png")
#users = utils.get_target_users("../../../../../Dataset/users_clusters/Coll_I.csv")
users = utils.get_target_users("../../../../../Dataset/target_users.csv")
with open("../../../../../Outputs/Coll_I.csv", 'w') as f:
f.write("user_id,item_list\n")
for user_id in users:
f.write(str(user_id) + ", " + utils.trim(recommender.recommend(user_id, at=10)) + "\n")
'''
URM_test = sps.csr_matrix(
sps.load_npz("../../../../Dataset/URM/data_test.npz"))
users = utils.get_target_users("../../../../Dataset/target_users.csv", seek=8)
URM = sps.csr_matrix(sps.load_npz("../../../../Dataset/URM/data_all.npz"))
validator = validate(URM_test, [10])
mauri_recsys = ItemKNNCFRecommender(URM)
mauri_recsys.fit(shrink=25,
topK=10,
similarity="jaccard",
feature_weighting="TF-IDF",
normalize=False)
print(evaluate.evaluate(users, mauri_recsys, URM_test, 10)["MAP"])
results = validator.evaluateRecommender(mauri_recsys)
print(results[0][10]["MAP"])
# SHRINK:10, K:100, SIMILARITY:asymmetric, FEATURE=none, NORM=False @ 0.023746