# ItemCF算法
# 统计所有电影之间相似度，存入数组中
def all_movies_sim(self):
arr_movies = list(np.array(self.users_movies).transpose()) # 计算电影相似度时，将原数组行转列
dp = calculate_sim(arr_movies, self.t)
return dp
# 相似电影推荐
def rec_movies(self, user, k):
m = self.users.index(user) # 用户user下标
looked = self.users_movies[m] # 用户m的电影队列
assert len(looked) == len(self.scores[m]), 'Unequal lengths'
# candidate[1]需要是一个数值越大排名越前的数组，所以取负
candidate = [[self.movies[i], -round(self.scores[m][i], 3)] for i in range(len(looked)) if looked[i] == 0]
res = top_k(candidate, k)
return res # 输出电影列表
# UserCF算法
def all_users_sim(self):
dp = calculate_sim(self.users_movies, self.t)
return dp
# 先输出最相似的k1个用户，再利用权重输出没看过电影的得分，输出前k2个电影作为推荐
def rec_movies(self, user, k1, k2):
m = self.users.index(user)
score = [0] * len(self.movies) # 初始分数为0
looked = self.users_movies[m] # 用户m的电影队列
tmp = []
if self.t == 'euc':
# 输出top k1个用户
candidate_user = [[self.users[i], -round(self.scores[m][i], 3)]
for i in range(len(self.users)) if self.users[i] != user]
res_user = top_k(candidate_user, k1)
# 权重计算为1/-res_user[i][0]，结果为正数且数字越大排名越前
tmp = [[self.users.index(res_user[i][1]), 1 / -res_user[i][0]] for i in range(len(res_user))]
if self.t == 'pea':
candidate_user = [[self.users[i], round(self.scores[m][i], 3)]
for i in range(len(self.users)) if self.users[i] != user]
res_user = top_k(candidate_user, k1)
# 因为结果为正数且数字越大排名越前，所以不做处理
tmp = [[self.users.index(res_user[i][1]), res_user[i][0]] for i in range(len(res_user))]
for x in tmp:
score = [score[i] + self.users_movies[x[0]][i] * x[1] for i in range(len(score))] # 利用权重计算总分数
candidate_score = [[self.movies[i], score[i]] for i in range(len(looked)) if looked[i] == 0]
# 输出top k2个电影
res = top_k(candidate_score, k2)
return res
