def worker(fold, n_users, n_items, dataset_dir):
traFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tra.txt'
trasR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, traFilePath),
binarize_threshold))
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape,
trasR.nnz, '%.2f' % (trasR.nnz / float(trasR.shape[0])))
tstFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tst.txt'
tstsR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, tstFilePath),
binarize_threshold))
sampler = Sampler(trasR, n_neg=negSample, batch_size=batch_size)
cml = CML(n_users, n_items, topN, split_method, eval_metrics, reg_cov,
margin, use_rank_weight, clip_norm, n_factors, batch_size)
scores = cml.train(fold + 1, trasR, tstsR, sampler)
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1),
','.join(['%s' % eval_metric for eval_metric in eval_metrics]) +
'@%d=' % (topN) + ','.join(['%.6f' % (score) for score in scores]))
cml.close()
return scores
def worker(fold, n_users, n_items, dataset_dir):
traFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tra.txt'
trasR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, traFilePath),
binarize_threshold))
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape,
trasR.nnz, '%.2f' % (trasR.nnz / float(trasR.shape[0])))
tstFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tst.txt'
tstsR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, tstFilePath),
binarize_threshold))
sampler = Sampler(trasR, gsize, negSample, batch_size)
gbprmf = GBPRMF(n_users, n_items, topN, rho, gsize, split_method,
eval_metrics, reg, n_factors, batch_size)
scores = gbprmf.train(fold + 1, trasR, tstsR, sampler)
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), 'gsize=', gsize,
'rho=', rho, 'reg=', reg)
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1),
','.join(['%s' % eval_metric for eval_metric in eval_metrics]) +
'@%d=' % (topN) + ','.join(['%.6f' % (score) for score in scores]))
gbprmf.close()
return scores
def worker(fold, n_users, n_items, dataset_dir):
traFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tra.txt'
trasR = loadSparseR(n_users, n_items, traFilePath)
print(
dataset_dir.split('/')[-2] + ':', trasR.shape, trasR.nnz,
'%.2f' % (trasR.nnz / float(trasR.shape[0])))
tra_tuple = np.array([(user, item, trasR[user, item])
for user, item in np.asarray(trasR.nonzero()).T
]) # triad
tstFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tst.txt'
tstsR = loadSparseR(n_users, n_items, tstFilePath)
tst_tuple = np.array([(user, item, tstsR[user, item])
for user, item in np.asarray(tstsR.nonzero()).T
]) # triad
sampler = Sampler(trasR=trasR, negRatio=.0, batch_size=batch_size)
svd = SVD(n_users, n_items, eval_metrics, range_of_ratings, reg, n_factors,
batch_size)
scores = svd.train(fold + 1, tra_tuple, tst_tuple, sampler)
print('fold=%d:' % fold,
','.join(['%s' % eval_metric for eval_metric in eval_metrics]), '=',
','.join(['%.6f' % (score) for score in scores]))
return scores
def worker(fold, n_users, n_items, dataset_dir):
traFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tra.txt'
trasR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, traFilePath),
binarize_threshold))
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape,
trasR.nnz, '%.2f' % (trasR.nnz / float(trasR.shape[0])))
tstFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tst.txt'
tstsR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, tstFilePath),
binarize_threshold))
prigp = PRIGP(n_users, n_items, topK, topN, split_method, eval_metrics,
alpha, reg, n_factors, batch_size)
scores = prigp.train(fold + 1, trasR, tstsR)
print('topK=', topK, 'alpha=', alpha, 'reg=', reg)
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1),
','.join(['%s' % eval_metric for eval_metric in eval_metrics]) +
'@%d=' % (topN) + ','.join(['%.6f' % (score) for score in scores]))
prigp.close()
return scores
def worker(fold, n_users, n_items, dataset_dir):
traFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tra.txt'
trasR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, traFilePath),
binarize_threshold))
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape,
trasR.nnz, '%.2f' % (trasR.nnz / float(trasR.shape[0])))
tstFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tst.txt'
tstsR = lil_matrix(
matBinarize(loadSparseR(n_users, n_items, tstFilePath),
binarize_threshold))
uicf = UserItemCF(n_users, n_items, topK, theta, topN, split_method,
eval_metrics)
scores = uicf.train(fold + 1, trasR, tstsR)
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1),
','.join(['%s' % eval_metric for eval_metric in eval_metrics]) +
'@%d=' % (topN) + ','.join(['%.6f' % (score) for score in scores]))
return scores
def worker(fold, n_A, n_B, dataset_dir, A_name, B_name):
A2B_trasR = lil_matrix(
loadSparseR(
n_A, n_B, dataset_dir + A_name + '2' + B_name + '_' +
str(fold + 1) + '_tra.txt'))
B2A_trasR = lil_matrix(
loadSparseR(
n_B, n_A, dataset_dir + B_name + '2' + A_name + '_' +
str(fold + 1) + '_tra.txt'))
trasR = lil_matrix((n_A + n_B, n_A + n_B))
trasR[:n_A, :n_B] = A2B_trasR
trasR[n_A:, n_B:] = B2A_trasR
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape,
trasR.nnz, '%.2f' % (trasR.nnz / float(trasR.shape[0])))
A2B_tstsR = lil_matrix(
loadSparseR(
n_A, n_B, dataset_dir + A_name + '2' + B_name + '_' +
str(fold + 1) + '_tst.txt'))
B2A_tstsR = lil_matrix(
loadSparseR(
n_B, n_A, dataset_dir + B_name + '2' + A_name + '_' +
str(fold + 1) + '_tst.txt'))
A2B_match_tstsR = A2B_tstsR.multiply(B2A_tstsR.T) > 0
sc = SocialCollab(topN, split_method, eval_metrics)
A2B_scores, B2A_scores = sc.train(fold + 1, A2B_trasR, B2A_trasR,
A2B_match_tstsR)
# bilateral_scores
tst_Anum, tst_Bnum = len(set(A2B_match_tstsR.nonzero()[0])), len(
set(A2B_match_tstsR.nonzero()[1]))
bi_scores = [
tst_Anum / float(tst_Anum + tst_Bnum) * A2B_scores[i] +
tst_Bnum / float(tst_Anum + tst_Bnum) * B2A_scores[i]
for i in range(len(eval_metrics))
]
# print results
print(','.join(['%s' % eval_metric
for eval_metric in eval_metrics]) + '@%d:' % (topN))
# bilateral
print(
dataset_dir.split('/')[-2] + '@%d: bi@%d=' % (fold + 1, topN) +
','.join(['%.6f' % (score) for score in bi_scores]))
return bi_scores
def split_tra_tst(usernum, itemnum, inFilePath, cv_fold):
sR = loadSparseR(usernum, itemnum, inFilePath)
useritemrating_tuples = np.array([
(user, item, sR[user, item])
for user, item in np.asarray(sR.nonzero()).T
])
np.random.shuffle(useritemrating_tuples)
fold_ins_num = int(len(useritemrating_tuples) / cv_fold)
outdir, filename = '/'.join(
inFilePath.split('/')[:-1]) + '/', inFilePath.split('/')[-1]
for cv_fold_ind in range(cv_fold):
tst_outFilePath = outdir + filename.replace(
'.', '_' + str(cv_fold_ind + 1) + '_tst.')
tst = useritemrating_tuples[cv_fold_ind *
fold_ins_num:(cv_fold_ind + 1) *
fold_ins_num, :]
saveTriads(tst, tst_outFilePath, isRatingInt=False)
tra_outFilePath = outdir + filename.replace(
'.', '_' + str(cv_fold_ind + 1) + '_tra.')
tra = np.concatenate([
useritemrating_tuples[:cv_fold_ind * fold_ins_num, :],
useritemrating_tuples[(cv_fold_ind + 1) * fold_ins_num:, :]
])
saveTriads(tra, tra_outFilePath, isRatingInt=False)
def worker(fold, n_users, n_items, dataset_dir):
traFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tra.txt'
trasR = lil_matrix(matBinarize(loadSparseR(n_users, n_items, traFilePath), binarize_threshold))
print(dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape, trasR.nnz,
'%.2f' % (trasR.nnz / float(trasR.shape[0])))
tstFilePath = dataset_dir + 'ratings__' + str(fold + 1) + '_tst.txt'
tstsR = lil_matrix(matBinarize(loadSparseR(n_users, n_items, tstFilePath), binarize_threshold))
sampler = Sampler(trasR, negRatio, batch_size)
wrmf = WRMF(n_users, n_items, topN, split_method, eval_metrics, weight, reg, n_factors, batch_size)
scores = wrmf.train(fold+1, trasR, tstsR, sampler)
print(dataset_dir.split('/')[-2] + '@%d:' % (fold + 1),'weight=',weight,'reg=',reg)
print('fold=%d:' % (fold+1), ','.join(['%s' % eval_metric for eval_metric in eval_metrics]), '=',
','.join(['%.6f' % (score) for score in scores]))
return scores
def worker(fold, n_A, n_B, dataset_dir, A_name, B_name):
A2B_trasR = lil_matrix(
loadSparseR(
n_A, n_B, dataset_dir + A_name + '2' + B_name + '_' +
str(fold + 1) + '_tra.txt'))
B2A_trasR = lil_matrix(
loadSparseR(
n_B, n_A, dataset_dir + B_name + '2' + A_name + '_' +
str(fold + 1) + '_tra.txt'))
trasR = lil_matrix((n_A + n_B, n_A + n_B))
trasR[:n_A, :n_B] = A2B_trasR
trasR[n_A:, n_B:] = B2A_trasR
print(
dataset_dir.split('/')[-2] + '@%d:' % (fold + 1), trasR.shape,
trasR.nnz, '%.2f' % (trasR.nnz / float(trasR.shape[0])))
A2B_tstsR = lil_matrix(
loadSparseR(
n_A, n_B, dataset_dir + A_name + '2' + B_name + '_' +
str(fold + 1) + '_tst.txt'))
B2A_tstsR = lil_matrix(
loadSparseR(
n_B, n_A, dataset_dir + B_name + '2' + A_name + '_' +
str(fold + 1) + '_tst.txt'))
rrtcf = RRK(n_A, n_B, topN, split_method, eval_metrics, reg, n_factors,
batch_size)
uni_scores, bi_scores = rrtcf.train(fold + 1, A2B_trasR, A2B_tstsR,
B2A_trasR, B2A_tstsR)
# print results
print(','.join(['%s' % eval_metric
for eval_metric in eval_metrics]) + '@%d:' % (topN))
# unilateral and bilateral
print(
dataset_dir.split('/')[-2] + '@%d: uni@%d=' % (fold + 1, topN) +
','.join(['%.6f' % (score) for score in uni_scores]) + ' bi@%d=' %
(topN) + ','.join(['%.6f' % (score) for score in bi_scores]))
return [uni_scores, bi_scores]