def main(k=100, user_sim=None, data_file=r'./data/movielens_data.pkl'):
train, test = pickle.load(open(data_file, 'rb+'))
train_itemset = stat_train_test_item(train)
test_itemset = stat_train_test_item(test)
trainset, test, item_len = utils.deal_train(data_file)
cf = UserCF(trainset, test, item_len, k=k)
degreedistrev = degree_item_map(cf)
Ndegree_items, Nitemset = getNdegree_items(degreedistrev, N=10)
start = time.time()
print('calculate similarity.')
if user_sim is None:
user_sim = cf.similarity()
else:
cf.user_sim = user_sim
print('calculate similarity finished. cost {:.2f}s'.format(time.time() -
start))
start = time.time()
recommend_score = cf.get_score(Nitemset)
print('calculate recommender score. cost {:.2f}s'.format(time.time() -
start))
user_degree = cf.cal_user_degree()
item_score = get_item_score(user_degree, recommend_score, item_len)
test_item_degree = get_test_degree(test)
print('start trend predict.')
corr_score = trend_predict(item_score,
Ndegree_items,
test_item_degree,
method='pearson')
print(corr_score)
return corr_score, recommend_score, user_sim
def main(our_lambda=1,data_file=r'./data/movielens_data.pkl',recommend_score_file=r'./temp/cf_score.pkl', recommend_score=None):
train, test = pickle.load(open(data_file, 'rb+'))
train_itemset = stat_train_test_item(train)
test_itemset = stat_train_test_item(test)
trainset, test, item_len = utils.deal_train(data_file)
cf = UserCF(trainset, test, item_len)
degreedistrev = degree_item_map(cf)
# get_item_degree_distribute(cf)
print('start cf train.')
if os.path.exists(recommend_score_file):
# 判断cf是否训练过
with open(recommend_score_file,'rb') as f:
recommend_score = pickle.load(f)
else:
recommend_score = cf.cf_train()
with open(recommend_score_file,'wb') as f:
pickle.dump(recommend_score, f)
# if recommend_score is None:
# recommend_score = cf.cf_train()
user_degree = cf.cal_user_degree()
item_score = get_item_score(user_degree, recommend_score, item_len, our_lambda=our_lambda)
Ndegree_items = getNdegree_items(degreedistrev, N=10)
test_item_degree = get_test_degree(test)
print('start trend predict.')
corr_score = trend_predict(item_score, Ndegree_items,test_item_degree,
train_itemset, test_itemset, method='pearson')
print(corr_score)
return corr_score, recommend_score
def degree_item_map(cf):
'''
建立degree-item倒排表,用于统计训练集中item的degree信息
'''
item_degrees = cf.cal_item_degree()
degreedistrev = {}
for iid, degree in item_degrees.items():
if degree not in degreedistrev:
degreedistrev[degree] = []
degreedistrev[degree].append(iid)
return degreedistrev
if __name__ == "__main__":
train_set, test, item_len = utils.deal_train(r'./data/movielens_data.pkl')
cf = UserCF(train_set, item_len)
cf.cal_user_degree()
cf.similarity()
# 定义一个浮点跳步器
def frange(x, y, jump):
while x < y:
yield x
x += jump
rec_score = cf.get_item_score()
degreedistrev = degree_item_map(cf)