通过TrustSVD算法进行基于矩阵分解的商品推-数据集

import numpy as np from numpy import inf def evaluation(test_R, test_mask_R, Estimated_R, num_test_ratings): pre_numerator = np.multiply((test_R - Estimated_R), test_mask_R) numerator = np.sum(np.square(pre_numerator)) denominator = num_test_ratings RMSE = np.sqrt(numerator / float(denominator)) pre_numeartor = np.multiply((test_R - Estimated_R), test_mask_R) numerator = np.sum(np.abs(pre_numeartor)) denominator = num_test_ratings MAE = numerator / float(denominator) pre_numeartor1 = np.sign(Estimated_R - 0.5) tmp_test_R = np.sign(test_R - 0.5) pre_numerator2 = np.multiply((pre_numeartor1 == tmp_test_R), test_mask_R) numerator = np.sum(pre_numerator2) denominator = num_test_ratings ACC = numerator / float(denominator) a = np.log(Estimated_R) b = np.log(1 - Estimated_R) a[a == -inf] = 0 b[b == -inf] = 0 tmp_r = test_R tmp_r = a * (tmp_r > 0) + b * (tmp_r == 0) tmp_r = np.multiply(tmp_r, test_mask_R) numerator = np.sum(tmp_r) denominator = num_test_ratings AVG_loglikelihood = numerator / float(denominator) return RMSE, MAE, ACC, AVG_loglikelihood def read_rating(datapath, dataname, num_users, num_items, random_seed): user_train_set = set() user_test_set = set() item_train_set = set() item_test_set = set() R = np.zeros((num_users, num_items)) mask_R = np.zeros((num_users, num_items)) C = np.zeros((num_users, num_items)) train_R = np.zeros((num_users, num_items)) test_R = np.zeros((num_users, num_items)) train_mask_R = np.zeros((num_users, num_items)) test_mask_R = np.zeros((num_users, num_items)) num_train_ratings = 0 num_test_ratings = 0 train_file_name = datapath + f'{dataname}.train' test_file_name = datapath + f'{dataname}.test' ''' load train fold ''' with open(train_file_name) as f1: lines = f1.readlines() for line in lines: user, item, voting = line.split("\t") user = int(user) item = int(item) voting = int(voting) if voting == -1: voting = 0 ''' Total ''' R[user, item] = voting mask_R[user, item] = 1 ''' Train ''' train_R[user, item] = int(voting) train_mask_R[user, item] = 1 C[user, item] = 1 user_train_set.add(user) item_train_set.add(item) num_train_ratings = num_train_ratings + 1 ''' load test fold ''' with open(test_file_name) as f2: lines = f2.readlines() for line in lines: user, item, voting = line.split("\t") user = int(user) item = int(item) voting = int(voting) if voting == -1: voting = 0 ''' Total ''' R[user, item] = voting mask_R[user, item] = 1 ''' Test ''' test_R[user, item] = int(voting) test_mask_R[user, item] = 1 user_test_set.add(user) item_test_set.add(item) num_test_ratings = num_test_ratings + 1 assert num_train_ratings == np.sum(train_mask_R) assert num_test_ratings == np.sum(test_mask_R) return train_R, train_mask_R, test_R, test_mask_R, num_train_ratings, num_test_ratings,\ user_train_set, item_train_set, user_test_set, item_test_set def read_trust(datapath, dataname, num_users): trust_matrix = np.zeros((num_users, num_users)) filename = datapath + f'{dataname}.trust' with open(filename) as f2: lines = f2.readlines() for line in lines: src, dst = line.split("\t") trust_matrix[int(src), int(dst)] = 1 return trust_matrix