def __init__(self, logger=None):
super().__init__()
self.logger = get_logger('item_cf')
if logger is not None:
self.logger = logger
self.user_score_dict = self.init_user_score()
# self.item_sim = self.item_similarity()
self.item_sim = self.item_similarity_best()
def __init__(self, datafile, logger=None):
self.logger = get_logger('user_cf_recommend')
if logger is not None:
self.logger = logger
self.datafile = datafile
self.data = self.load_data()
self.train_data, self.test_data = self.split_data(3, 47)
self.user_sim = self.user_similarity_best()
def __init__(self, datafile, ratio, logger=None):
self.logger = get_logger('user_cf_recommend')
if logger is not None:
self.logger = logger
# 原始数据路径文件
self.datafile = datafile
# 测试集与训练集的比例
self.ratio = ratio
self.item_sim_file = 'data/item_sim.json'
self.data = self.load_data()
self.train_data, self.test_data = self.split_data(3, 47)
self.item_sim = self.item_similarity_best()
"""
score = 0.0
for item1 in self.item_sim[item].keys():
if item1 != item:
score += self.item_sim[item][item1] * self.user_score_dict[
user][item1]
return score
def recommend(self, user):
"""
为用户推荐物品
:param user:
:return:
"""
# 计算user未评分item的可能评分
user_item_score_dict = dict()
for item in self.user_score_dict[user].keys():
if self.user_score_dict[user][item] <= 0:
user_item_score_dict[item] = self.pre_user_item_score(
user, item)
return user_item_score_dict
if __name__ == '__main__':
logger = get_logger('item_cf')
ib = ItemCF(logger=logger)
logger.info(ib.recommend('C'))
logger.info(ib.recommend('A'))
计算准确率
:param k:
:param nitems:
:return:
"""
self.logger.info('开始计算准确率...')
hit = 0
precision = 0
for user in self.test_data.keys():
u_items = self.test_data.get(user, {})
result = self.recommend(user, k=k, nitems=nitems)
for item, rate in result.items():
if item in u_items:
hit += 1
precision += nitems
return hit / (precision * 1.0)
if __name__ == '__main__':
logger = get_logger('item_cf_movie_recommend')
ib = ItemCFRec(datafile='../../data/ml-1m/ratings.dat',
ratio=[1, 9],
logger=logger)
logger.info(f'用户1进行推荐的结果如下: {ib.recommend("1")}')
for k in range(5, 11, 1):
for nitems in range(10, 22, 2):
precision = ib.precision(k=k, nitems=nitems)
logger.info(f'k = {k}, nitems={nitems}, precision is {precision}')
precision = 0
for user in self.train_data.keys():
# tu: 用户实际评分的电影ID
tu = self.test_data.get(user, {})
# rank: 推荐系统推荐的电影ID
rank = self.recommend(user, k=k, nitems=nitems)
for item, rate in rank.items():
if item in tu:
hit += 1
precision += nitems
self.logger.info(f'hit: {hit}')
self.logger.info(f'precision: {precision}')
return hit / (precision * 1.0)
if __name__ == '__main__':
logger = get_logger('user_cf_recommend')
cf = UserCFRec('../../data/ml-1m/ratings.dat', logger=logger)
result = cf.recommend('1')
logger.info(f'user "1" recommend result is {result}')
precision = cf.precision()
logger.info(f'k=8, n=10, precision is {precision}')
for k in range(5, 10, 1):
for nitems in range(10, 20, 2):
precision = cf.precision(k=k, nitems=nitems)
logger.info(f'k = {k}, nitems={nitems}, precision is {precision}')