def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
user_k = int(args[3])
try:
top_n = int(args[4])
except:
top_n = 500 #输出top500的推荐到文件
#Filepath config
item_tag_file = './song_dataset/mid_data/song_tag_distribution.json'
user_tag_file = './song_dataset/mid_data/user_tag_distribution_%s_%s.json'%(set_level,train_prob)
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
user_sim_file = './song_dataset/mid_data/user_similarity_withTag_%s_%s.json'%(set_level,train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
#Initiate recommender
recommender = UserTagCF()
recommender.build_userTagDistribution(dataset.train_data,item_tag_file,user_tag_file)
recommender.build_user_similarity(dataset.train_data,user_sim_file,top_user_k=1000) #保留用户的top1000个最相近的用户
#Recommendation
recommender.recommend(dataset.train_data,user_k=user_k,top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
user_k = int(args[3])
try:
top_n = int(args[4])
except:
top_n = 500
#File path config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json' % (
set_level, train_prob) # user-user simiarity matrix
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s" % (dataset.cost_time))
#Initiate Recommender
recommender = UserCF()
recommender.build_user_similarity(
dataset.train_data, user_sim_file,
top_user_k=1000) #Top_user_k represent keep top k sim_user to file
#Recommendation
recommender.recommend(dataset.train_data, user_k=user_k, top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s" %
(train_prob, user_k, top_n, recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
user_k = int(args[3])
try:
top_n = int(args[4])
except:
top_n = 500
#File path config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json'%(set_level,train_prob) # user-user simiarity matrix
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
#Initiate Recommender
recommender = UserCF()
recommender.build_user_similarity(dataset.train_data,user_sim_file,top_user_k=1000) #Top_user_k represent keep top k sim_user to file
#Recommendation
recommender.recommend(dataset.train_data,user_k=user_k,top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
top_n = int(args[3])
#Filepath config
item_tag_file = './song_dataset/mid_data/song_tag_distribution.json'
#item_tag_file = './song_dataset/mid_data/song_tag_dist_with_singer.json'
user_tag_file = './song_dataset/mid_data/user_tag_distribution_%s_%s.json'%(set_level,train_prob)
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
user_sim_file = './song_dataset/mid_data/user_similarity_withTag_%s_%s.json'%(set_level,train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
print "DataForTrain: %s"%(train_file)
print "DataForTest: %s"%(test_file)
print "Dataset train_set info: %s"%(dataset.get_train_info())
print "Dataset test_set info: %s"%(dataset.get_test_info())
#Record best scores
best_f_score = {'f_score':0}
best_precision = {'precision':0}
best_recall = {'recall':0}
#Initiate recommender
recommender = UserTagCF()
recommender.build_userTagDistribution(dataset.train_data,item_tag_file,user_tag_file)
recommender.build_user_similarity(dataset.train_data,user_sim_file,top_user_k=1000)
#Recommendation
for user_k in [5]+range(10,101,10):
recommender.recommend(dataset.train_data,user_k=user_k,top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
scores = recommender.score(dataset.test_data,len(dataset.all_songs))
print "User_k:%s\tTop_n:%s\tScores:%s"%(user_k,top_n,scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['user_k'] = user_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['user_k']=user_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['user_k']=user_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s"%(best_f_score)
print "Best_Precision: %s"%(best_precision)
print "Best_Recall: %s"%(best_recall)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
topic_num = int(args[3])
recommend_job = args[4]
user_k = int(args[5])
try:
top_n = int(args[6])
except:
top_n = 500
#Log config
log_file = './log/hybirdModel_%s_%s_%s_%s_%s.log'%(set_level,train_prob,topic_num,recommend_job,top_n)
logging.basicConfig(level=logging.INFO,format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s',filename=log_file)
#Filepath config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json'%(set_level,train_prob)
userTag_sim_file = './song_dataset/mid_data/user_similarity_withTag_%s_%s.json'%(set_level,train_prob)
userLDA_sim_file = './song_dataset/mid_data/user_sim_with_lda_%s_%s_%s.json'%(set_level,train_prob,topic_num)
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
#Data Preparation
items_tag_dict = {}
users_tag_dict = {}
if recommend_job in ('mix_result_reorder','mix_sim_reorder'):
items_tag_dict = load_tag_distribution('./song_dataset/mid_data/song_tag_distribution.json') #Load item_tag_distrib
user_tag_file = './song_dataset/mid_data/user_tag_distribution_%s_%s.json'%(set_level,train_prob)
users_tag_dict = load_tag_distribution(user_tag_file)
#Initiate Hybird-Model
recommender = HybirdModel()
if recommend_job in ('mix_sim','mix_sim_reorder'):
recommender.hybird_user_sim(dataset.train_data,userTag_sim_file,userLDA_sim_file,theta=0.45)
elif recommend_job in ('mix_result','mix_result_reorder'):
recommender.userTag.load_user_similarity(userTag_sim_file,norm=1)
recommender.userLda.load_user_similarity(userLDA_sim_file,norm=1)
if recommend_job == 'mix_sim':
recommender.recommend(dataset.train_data,users_tag_dict,items_tag_dict,user_k,top_n,reorder=0)
elif recommend_job == 'mix_sim_reorder':
recommender.recommend(dataset.train_data,users_tag_dict,items_tag_dict,user_k,top_n,reorder=1)
elif recommend_job == 'mix_result':
recommender.hybird_recommend_result(dataset.train_data,user_k,top_n)
elif recommend_job == 'mix_result_reorder':
recommender.hybird_result_withReorder(dataset.train_data,users_tag_dict,items_tag_dict,user_k,top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
dataset = BaseDataSet()
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level,type,train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s",dataset.cost_time)
recommender = RandomSelect()
recommender.recommend(dataset.train_data,list(dataset.all_songs),500)
logging.info("Train_prob:%s cost:%s"%(train_prob,recommender.get_time()))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
dataset = BaseDataSet()
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_levle,type,train_prob
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s", dataset.cost_time)
#Initiate Recommender
recommender = Popularity()
recommender.recommend(dataset.train_data, 500)
logging.info("Train_prob:%s cost:%s" % (train_prob, recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
dataset = BaseDataSet()
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_levle,type,train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s",dataset.cost_time)
#Initiate Recommender
recommender = Popularity()
recommender.recommend(dataset.train_data,500)
logging.info("Train_prob:%s cost:%s"%(train_prob,recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
e_type = args[3] #Experiment type: song or playlist
dataset = BaseDataSet()
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_levle,type,train_prob
if e_type == 'playlist':
file_template = './pl_dataset/user_playlist_%s_%s_%s' #set_levle,type,train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s",dataset.cost_time)
print "DataForTrain: %s"%(train_file)
print "DataForTest: %s"%(test_file)
print "Dataset train_set info: %s"%(dataset.get_train_info())
print "Dataset test_set info: %s"%(dataset.get_test_info())
#Record best scores
best_f_score = {'f_score':0}
best_precision = {'precision':0}
best_recall = {'recall':0}
#Initiate Recommender
recommender = Popularity()
for i in [1,50,100,150,200]:
recommender.recommend(dataset.train_data,i)
logging.info("Train_prob:%s Recommend Top_n:%s cost:%s"%(train_prob,i,recommender.cost_time))
#logging.info("Top_10_song:%s"%(recommender.get_poplist(10)))
scores = recommender.score(dataset.test_data)
print "Top_n:%s\tScores:%s"%(i,scores)
#Find best scores
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['top_n'] = i
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['top_n'] = i
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['top_n'] = i
print "Best_F_Score: %s"%(best_f_score)
print "Best_Precision: %s"%(best_precision)
print "Best_Recall: %s"%(best_recall)
def get_lda_topics(args):
set_level = args[0]
train_prob = args[1]
topic_num = int(args[2])
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
recommender = UserLDA()
recommender.build_model(dataset.train_data,topic_num)
for idx,distrib in enumerate(recommender.model.print_topics(1000)):
dist0 = distrib.split()[0].split('*')[0]
if float(dist0) > 0:
print "Topic#%s\t%s"%(idx,distrib)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
topic_num = int(args[3])
user_k = int(args[4])
try:
top_n = int(args[5])
except:
top_n = 500
#Log-Config
logfile = './log/userLDA_%s_%s_%s.log' % (set_level, train_prob, topic_num)
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s',
filename=logfile)
#logging.basicConfig(level=logging.INFO,format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s')
#File path config
user_sim_file = './song_dataset/mid_data/user_sim_with_lda_%s_%s_%s.json' % (
set_level, train_prob, topic_num)
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s" % (dataset.cost_time))
#Initiate Recommender
recommender = UserLDA()
recommender.build_user_similarity(user_sim_file,
dataset.train_data,
topic_num=topic_num,
top_user_k=300)
#Recommendation
recommender.recommend(dataset.train_data, user_k=user_k, top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s" %
(train_prob, user_k, top_n, recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
kn = int(args[3])
#File path config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s" % (dataset.cost_time))
#Initiate Recommender
recommender = SVDModel()
recommender.build_model(dataset.train_data, kn)
recommender.recommend(dataset.train_data)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
kn = int(args[3])
#File path config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
#Initiate Recommender
recommender = SVDModel()
recommender.build_model(dataset.train_data,kn)
recommender.recommend(dataset.train_data)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
top_n = int(args[3])
#File path config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json'%(set_level,train_prob) # user-user simiarity matrix
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
print "DataForTrain: %s"%(train_file)
print "DataForTest: %s"%(test_file)
print "Dataset train_set info: %s"%(dataset.get_train_info())
print "Dataset test_set info: %s"%(dataset.get_test_info())
#Record best scores
best_f_score = {'f_score':0}
best_precision = {'precision':0}
best_recall = {'recall':0}
#Initiate Recommender
recommender = UserCF()
recommender.build_user_similarity(dataset.train_data,user_sim_file,top_user_k=1000) #Top_user_k represent keep top k sim_user to file
#Recommendation
for user_k in [5]+range(10,101,10):
recommender.recommend(dataset.train_data,user_k=user_k,top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
scores = recommender.score(dataset.test_data,len(dataset.all_songs))
print "User_k:%s\tTop_n:%s\tScores:%s"%(user_k,top_n,scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['user_k'] = user_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['user_k']=user_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['user_k']=user_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s"%(best_f_score)
print "Best_Precision: %s"%(best_precision)
print "Best_Recall: %s"%(best_recall)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
e_type = args[3] #Experiment type: song or playlist
dataset = BaseDataSet()
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_levle,type,train_prob
if e_type == 'playlist':
file_template = './pl_dataset/user_playlist_%s_%s_%s' #set_levle,type,train_prob
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s", dataset.cost_time)
print "DataForTrain: %s" % (train_file)
print "DataForTest: %s" % (test_file)
print "Dataset train_set info: %s" % (dataset.get_train_info())
print "Dataset test_set info: %s" % (dataset.get_test_info())
#Record best scores
best_f_score = {'f_score': 0}
best_precision = {'precision': 0}
best_recall = {'recall': 0}
#Initiate Recommender
recommender = Popularity()
for i in [1, 50, 100, 150, 200]:
recommender.recommend(dataset.train_data, i)
logging.info("Train_prob:%s Recommend Top_n:%s cost:%s" %
(train_prob, i, recommender.cost_time))
#logging.info("Top_10_song:%s"%(recommender.get_poplist(10)))
scores = recommender.score(dataset.test_data)
print "Top_n:%s\tScores:%s" % (i, scores)
#Find best scores
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['top_n'] = i
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['top_n'] = i
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['top_n'] = i
print "Best_F_Score: %s" % (best_f_score)
print "Best_Precision: %s" % (best_precision)
print "Best_Recall: %s" % (best_recall)
def __init__(self):
BaseDataSet.__init__(self)
self.playlists = []
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
topic_num = int(args[3])
top_n = int(args[4])
e_type = args[5] #e_type: song or playlist
#Log-Config
logfile = './log/userLDA_%s_%s_%s.log'%(set_level,train_prob,topic_num)
logging.basicConfig(level=logging.INFO,format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s',filename=logfile,filemode='w')
#logging.basicConfig(level=logging.INFO,format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s')
#File path config
user_sim_file = './song_dataset/mid_data/user_sim_with_lda_%s_%s_%s_new.json'%(set_level,train_prob,topic_num)
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
if e_type == 'playlist':
user_sim_file = './pl_dataset/mid_data/user_sim_with_lda_%s_%s_%s.json'%(set_level,train_prob,topic_num)
file_template = './pl_dataset/user_playlist_%s_%s_%s' #set_level, type, train_prob
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
print "DataForTrain: %s"%(train_file)
print "DataForTest: %s"%(test_file)
print "Dataset train_set info: %s"%(dataset.get_train_info())
print "Dataset test_set info: %s"%(dataset.get_test_info())
#Record best scores
best_f_score = {'f_score':0}
best_precision = {'precision':0}
best_recall = {'recall':0}
#Initiate Recommender
recommender = UserLDA()
recommender.build_user_similarity(user_sim_file,dataset.train_data,topic_num=topic_num, top_user_k=1000)
#Recommendation
for user_k in [5]+range(10,101,10):
recommender.recommend(dataset.train_data,user_k=user_k,top_n=top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
scores = recommender.score(dataset.test_data,len(dataset.all_songs))
print "User_k:%s\tTop_n:%s\tScores:%s"%(user_k,top_n,scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['user_k'] = user_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['user_k']=user_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['user_k']=user_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s"%(best_f_score)
print "Best_Precision: %s"%(best_precision)
print "Best_Recall: %s"%(best_recall)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
top_n = int(args[3])
e_type = args[4] #Experiment type: song or playlist
#Filepath config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
item_sim_file = './song_dataset/mid_data/item_similarity_%s_%s.json' % (
set_level, train_prob)
if e_type == 'playlist':
file_template = './pl_dataset/user_playlist_%s_%s_%s' #set_num,type,train_prob
item_sim_file = './pl_dataset/mid_data/item_similarity_%s_%s.json' % (
set_level, train_prob)
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s" % (dataset.cost_time))
print "DataForTrain: %s" % (train_file)
print "DataForTest: %s" % (test_file)
print "Dataset train_set info: %s" % (dataset.get_train_info())
print "Dataset test_set info: %s" % (dataset.get_test_info())
#Record best scores
best_f_score = {'f_score': 0}
best_precision = {'precision': 0}
best_recall = {'recall': 0}
#Initiate recommender
itemCF_recommender = ItemCF()
if os.path.exists(item_sim_file):
logging.info("File %s exists, loading item similarity matrix" %
(item_sim_file))
itemCF_recommender.load_item_similarity(item_sim_file)
logging.info("Load item_similarity cost: %s" %
(itemCF_recommender.cost_time))
else:
logging.info("File %s doesn't exist, building item similarity matrix" %
(item_sim_file))
itemCF_recommender.build_item_similarity(dataset.train_data,
item_sim_file)
logging.info("Load item_similarity cost: %s" %
(itemCF_recommender.cost_time))
#Recommendation
for item_k in range(20, 100):
itemCF_recommender.recommend(dataset.train_data,
item_k=item_k,
top_n=top_n)
logging.info("Train_prob:%s Item_k:%s Top_n:%s Cost:%s" %
(train_prob, item_k, top_n, itemCF_recommender.cost_time))
scores = itemCF_recommender.score(dataset.test_data)
print "Item_k:%s\tTop_n:%s\tScores:%s" % (item_k, top_n, scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['item_k'] = item_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['item_k'] = item_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['item_k'] = item_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s" % (best_f_score)
print "Best_Precision: %s" % (best_precision)
print "Best_Recall: %s" % (best_recall)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
hybird_type = args[3]
recommend_job = args[4]
user_k = int(args[5])
top_n = int(args[6])
if hybird_type == 'lda':
topic_num = int(args[7])
#Log config
log_file = './log/ubase_hybirdModel_%s_%s_%s_%s.log' % (
set_level, train_prob, recommend_job, top_n)
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s',
filename=log_file,
filemode='w')
#logging.basicConfig(level=logging.INFO,format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s')
#Filepath config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json' % (
set_level, train_prob) #Similarity file of userCF
if hybird_type == 'tag':
userTag_sim_file = './song_dataset/mid_data/user_similarity_withTag_%s_%s.json' % (
set_level, train_prob)
if hybird_type == 'lda':
userLDA_sim_file = './song_dataset/mid_data/user_sim_with_lda_%s_%s_%s.json' % (
set_level, train_prob, topic_num)
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s" % (dataset.cost_time))
#Data Preparation
items_tag_dict = {}
users_tag_dict = {}
if recommend_job in ('mix_result_reorder', 'mix_sim_reorder'):
items_tag_dict = load_tag_distribution(
'./song_dataset/mid_data/song_tag_distribution.json'
) #Load item_tag_distrib
user_tag_file = './song_dataset/mid_data/user_tag_distribution_%s_%s.json' % (
set_level, train_prob)
users_tag_dict = load_tag_distribution(user_tag_file)
#Initiate Hybird-Model
recommender = HybirdModel_UB()
if recommend_job in ('mix_sim', 'mix_sim_reorder'):
if hybird_type == 'tag':
recommender.hybird_user_sim(dataset.train_data,
user_sim_file,
userTag_sim_file,
hybird_type='tag',
theta=0.8,
mix_type=0)
elif hybird_type == 'lda':
recommender.hybird_user_sim(dataset.train_data,
user_sim_file,
userLDA_sim_file,
hybird_type='lda',
theta=0.9,
mix_type=0)
elif recommend_job in ('mix_result', 'mix_result_reorder'):
if hybird_type == 'tag':
recommender.userTag.load_user_similarity(userTag_sim_file, norm=1)
elif hybird_type == 'lda':
recommender.userLda.load_user_similarity(userLDA_sim_file, norm=1)
if recommend_job == 'mix_sim':
recommender.recommend(dataset.train_data,
users_tag_dict,
items_tag_dict,
user_k,
top_n,
reorder=0)
elif recommend_job == 'mix_sim_reorder':
recommender.recommend(dataset.train_data,
users_tag_dict,
items_tag_dict,
user_k,
top_n,
reorder=1)
elif recommend_job == 'mix_result':
recommender.hybird_recommend_result(dataset.train_data, user_k, top_n)
elif recommend_job == 'mix_result_reorder':
recommender.hybird_result_withReorder(dataset.train_data,
users_tag_dict, items_tag_dict,
user_k, top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s" %
(train_prob, user_k, top_n, recommender.cost_time))
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
topic_num = int(args[3])
top_n = int(args[4])
recommend_job = args[5]
#Log config
log_file = './log/hybirdModel_%s_%s_%s_%s_%s.log'%(set_level,train_prob,topic_num,recommend_job,top_n)
logging.basicConfig(level=logging.INFO,format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s',filename=log_file,filemode='w')
#Filepath config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json'%(set_level,train_prob)
userTag_sim_file = './song_dataset/mid_data/user_similarity_withTag_%s_%s.json'%(set_level,train_prob)
userLDA_sim_file = './song_dataset/mid_data/user_sim_with_lda_%s_%s_%s.json'%(set_level,train_prob,topic_num)
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
print "DataForTrain: %s"%(train_file)
print "DataForTest: %s"%(test_file)
print "Dataset train_set info: %s"%(dataset.get_train_info())
print "Dataset test_set info: %s"%(dataset.get_test_info())
#Record best scores
best_f_score = {'f_score':0}
best_precision = {'precision':0}
best_recall = {'recall':0}
#Data Preparation
items_tag_dict = {}
users_tag_dict = {}
if recommend_job in ('mix_result_reorder','mix_sim_reorder'):
items_tag_dict = load_tag_distribution('./song_dataset/mid_data/song_tag_distribution.json') #Load item_tag_distrib
user_tag_file = './song_dataset/mid_data/user_tag_distribution_%s_%s.json'%(set_level,train_prob)
users_tag_dict = load_tag_distribution(user_tag_file)
#Initiate Hybird-Model
recommender = HybirdModel()
if recommend_job in ('mix_sim','mix_sim_reorder'):
recommender.hybird_user_sim(dataset.train_data,userTag_sim_file,userLDA_sim_file,theta=0.45)
elif recommend_job in ('mix_result','mix_result_reorder'):
recommender.userTag.load_user_similarity(userTag_sim_file,norm=1)
recommender.userLda.load_user_similarity(userLDA_sim_file,norm=1)
for user_k in [5]+range(10,101,10):
if recommend_job == 'mix_sim':
recommender.recommend(dataset.train_data,users_tag_dict,items_tag_dict,user_k,top_n,reorder=0)
elif recommend_job == 'mix_sim_reorder':
recommender.recommend(dataset.train_data,users_tag_dict,items_tag_dict,user_k,top_n,reorder=1)
elif recommend_job == 'mix_result':
recommender.hybird_recommend_result(dataset.train_data,user_k,top_n)
elif recommend_job == 'mix_result_reorder':
recommender.hybird_result_withReorder(dataset.train_data,users_tag_dict,items_tag_dict,user_k,top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s"%(train_prob,user_k,top_n,recommender.cost_time))
scores = recommender.score(dataset.test_data,len(dataset.all_songs))
print "User_k:%s\tTop_n:%s\tScores:%s"%(user_k,top_n,scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['user_k'] = user_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['user_k']=user_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['user_k']=user_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s"%(best_f_score)
print "Best_Precision: %s"%(best_precision)
print "Best_Recall: %s"%(best_recall)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
top_n = int(args[3])
e_type = args[4] #Experiment type: song or playlist
#Filepath config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_num,type,train_prob
item_sim_file = './song_dataset/mid_data/item_similarity_%s_%s.json'%(set_level,train_prob)
if e_type == 'playlist':
file_template = './pl_dataset/user_playlist_%s_%s_%s' #set_num,type,train_prob
item_sim_file = './pl_dataset/mid_data/item_similarity_%s_%s.json'%(set_level,train_prob)
train_file = file_template%(set_level,'train',train_prob)
test_file = file_template%(set_level,'test',train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file,test_file)
logging.info("Build dataset cost:%s"%(dataset.cost_time))
print "DataForTrain: %s"%(train_file)
print "DataForTest: %s"%(test_file)
print "Dataset train_set info: %s"%(dataset.get_train_info())
print "Dataset test_set info: %s"%(dataset.get_test_info())
#Record best scores
best_f_score = {'f_score':0}
best_precision = {'precision':0}
best_recall = {'recall':0}
#Initiate recommender
itemCF_recommender = ItemCF()
if os.path.exists(item_sim_file):
logging.info("File %s exists, loading item similarity matrix"%(item_sim_file))
itemCF_recommender.load_item_similarity(item_sim_file)
logging.info("Load item_similarity cost: %s"%(itemCF_recommender.cost_time))
else:
logging.info("File %s doesn't exist, building item similarity matrix"%(item_sim_file))
itemCF_recommender.build_item_similarity(dataset.train_data,item_sim_file)
logging.info("Load item_similarity cost: %s"%(itemCF_recommender.cost_time))
#Recommendation
for item_k in range(20,100):
itemCF_recommender.recommend(dataset.train_data,item_k=item_k,top_n=top_n)
logging.info("Train_prob:%s Item_k:%s Top_n:%s Cost:%s"%(train_prob,item_k,top_n,itemCF_recommender.cost_time))
scores = itemCF_recommender.score(dataset.test_data)
print "Item_k:%s\tTop_n:%s\tScores:%s"%(item_k,top_n,scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['item_k'] = item_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['item_k']=item_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['item_k']=item_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s"%(best_f_score)
print "Best_Precision: %s"%(best_precision)
print "Best_Recall: %s"%(best_recall)
def main():
args = sys.argv
set_level = args[1]
train_prob = args[2]
topic_num = int(args[3])
top_n = int(args[4])
recommend_job = args[5]
#Log config
log_file = './log/hybirdModel_%s_%s_%s_%s_%s.log' % (
set_level, train_prob, topic_num, recommend_job, top_n)
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s %(levelname)s %(funcName)s %(lineno)d %(message)s',
filename=log_file,
filemode='w')
#Filepath config
file_template = './song_dataset/user_dataset_%s_%s_%s' #set_level, type, train_prob
user_sim_file = './song_dataset/mid_data/user_sim_%s_%s.json' % (
set_level, train_prob)
userTag_sim_file = './song_dataset/mid_data/user_similarity_withTag_%s_%s.json' % (
set_level, train_prob)
userLDA_sim_file = './song_dataset/mid_data/user_sim_with_lda_%s_%s_%s.json' % (
set_level, train_prob, topic_num)
train_file = file_template % (set_level, 'train', train_prob)
test_file = file_template % (set_level, 'test', train_prob)
#Build dataset
dataset = BaseDataSet()
dataset.build_data(train_file, test_file)
logging.info("Build dataset cost:%s" % (dataset.cost_time))
print "DataForTrain: %s" % (train_file)
print "DataForTest: %s" % (test_file)
print "Dataset train_set info: %s" % (dataset.get_train_info())
print "Dataset test_set info: %s" % (dataset.get_test_info())
#Record best scores
best_f_score = {'f_score': 0}
best_precision = {'precision': 0}
best_recall = {'recall': 0}
#Data Preparation
items_tag_dict = {}
users_tag_dict = {}
if recommend_job in ('mix_result_reorder', 'mix_sim_reorder'):
items_tag_dict = load_tag_distribution(
'./song_dataset/mid_data/song_tag_distribution.json'
) #Load item_tag_distrib
user_tag_file = './song_dataset/mid_data/user_tag_distribution_%s_%s.json' % (
set_level, train_prob)
users_tag_dict = load_tag_distribution(user_tag_file)
#Initiate Hybird-Model
recommender = HybirdModel()
if recommend_job in ('mix_sim', 'mix_sim_reorder'):
recommender.hybird_user_sim(dataset.train_data,
userTag_sim_file,
userLDA_sim_file,
theta=0.45)
elif recommend_job in ('mix_result', 'mix_result_reorder'):
recommender.userTag.load_user_similarity(userTag_sim_file, norm=1)
recommender.userLda.load_user_similarity(userLDA_sim_file, norm=1)
for user_k in [5] + range(10, 101, 10):
if recommend_job == 'mix_sim':
recommender.recommend(dataset.train_data,
users_tag_dict,
items_tag_dict,
user_k,
top_n,
reorder=0)
elif recommend_job == 'mix_sim_reorder':
recommender.recommend(dataset.train_data,
users_tag_dict,
items_tag_dict,
user_k,
top_n,
reorder=1)
elif recommend_job == 'mix_result':
recommender.hybird_recommend_result(dataset.train_data, user_k,
top_n)
elif recommend_job == 'mix_result_reorder':
recommender.hybird_result_withReorder(dataset.train_data,
users_tag_dict,
items_tag_dict, user_k,
top_n)
logging.info("Train_prob:%s User_k:%s Top_n:%s cost:%s" %
(train_prob, user_k, top_n, recommender.cost_time))
scores = recommender.score(dataset.test_data, len(dataset.all_songs))
print "User_k:%s\tTop_n:%s\tScores:%s" % (user_k, top_n, scores)
#Find Best Score
if scores['f_score'] > best_f_score['f_score']:
best_f_score = scores
best_f_score['user_k'] = user_k
best_f_score['top_n'] = top_n
if scores['precision'] > best_precision['precision']:
best_precision = scores
best_precision['user_k'] = user_k
best_precision['top_n'] = top_n
if scores['recall'] > best_recall['recall']:
best_recall = scores
best_recall['user_k'] = user_k
best_recall['top_n'] = top_n
print "Best_F_Score: %s" % (best_f_score)
print "Best_Precision: %s" % (best_precision)
print "Best_Recall: %s" % (best_recall)