import read_file as rf
import numpy as np
import computing as compute
import time
def get_predict_rui_ruj(uu, ii, jj):
rui = np.dot(U[uu], V[ii]) + bi[ii]
ruj = np.dot(U[uu], V[jj]) + bi[jj]
return rui, ruj
train_records, yui, iu, i_te_u, u_test = rf.read_train_test()
n = 943
m = 1682
K = 5
d = 20
global_avg = 0.0
numerator = 0.0
denominator = 0.0
for u in range(n):
for i in range(m):
numerator += yui[u][i]
denominator = m * n
global_avg = float(numerator) / denominator
# initialization of model parameters
bi = np.zeros(m, dtype=float)
import read_file as rf import computing as cp data = rf.read_train_test() # 从文件中读取 u1.base u1.test 得到两个二维数组 train_data = data[0] test_data = data[1] # 计算 各个用户 的本身的所有打分的平均值 得到 aver_u_to_i 列表 aver_u_to_i = cp.compute_aver_u_to_i(train_data) # 计算 两两用户 之间的 相似度 二维数组 cp.compute_Skj(train_data, aver_u_to_i)
def get_uu_i(uu, ii):
i_set = {ii}
iu_minus_i = list(set(iu[uu]) - i_set)
uu_i = np.zeros(d, dtype=float)
if len(iu_minus_i) <= 0:
return iu_minus_i, uu_i
for iii in iu_minus_i:
uu_i = uu_i + W[iii]
uu_i = uu_i / math.pow(len(iu_minus_i), alpha)
return iu_minus_i, uu_i
# observed_records, yui, iu, i_te_u, u_test
observed_records, all_unobserved_records, yui, iu, i_te_u, u_test = rf.read_train_test(
)
n = 943
m = 1682
K = 5
d = 20
global_avg = 0.0
numerator = 0.0
denominator = 0.0
for u in range(n):
for i in range(m):
numerator += yui[u][i]
denominator = m * n
global_avg = float(numerator) / denominator
import numpy as np
import read_file as rf
import perform_evaluate as pe
# user_p_i, item_ped_u, i_te_u, u_test, yui
user_p_i, item_ped_u, i_te_u, u_test, yui = rf.read_train_test()
n = 943
m = 1682
K = 50
Swu = np.zeros((n, n), dtype=np.float)
max_swu = np.zeros(n, dtype=np.float)
# 计算相似度 Swu
for user_w in range(n):
for user_u in range(n):
iw = user_p_i[user_w]
iu = user_p_i[user_u]
numerator = len(list(set(iw) & set(iu)))
denominator = len(list(set(iw) | set(iu)))
if denominator != 0:
Swu[user_w][user_u] = float(numerator) / denominator
if max_swu[user_w] < Swu[user_w][user_u]:
max_swu[user_w] = Swu[user_w][user_u]
# 正则化
for user_w in range(n):
import read_file as rf
import evaluation_etrics as ee
import numpy as np
ui_train_pairs, yui, Iu, I, U, ui_test_pair, I_te_u, U_test = rf.read_train_test(
)
n = 943
m = 1682
K = 5
global_avg = 0.0
numerator = 0.0
denominator = 0.0
for u in range(n):
for i in range(m):
numerator += yui[u][i]
denominator = n * m
global_avg = numerator / denominator
# recommendation list of items for user u a dict for every element, only have k items
# u from U_test
I_re_u = {}
for u in U_test:
I_re_u[u] = []
# recommendation list of items for user u, for every element, have more than k items
I_re_u_mk = {}
for u in U_test: