def get_performance(user_pos_test, r, auc, Ks):
'''
:param user_pos_test: user 测试集中真实交互的item
:param r: r = [1,0,1] 表示预测TOP-K是否命中
:param auc: auc =0 标量
:param Ks: TOP-K
:return:
'''
precision, recall, ndcg, hit_ratio, MAP = [], [], [], [], []
for K in Ks:
precision.append(metrics.precision_at_k(r, K))
recall.append(metrics.recall_at_k(r, K, len(user_pos_test)))
ndcg.append(metrics.ndcg_at_k(r, K))
hit_ratio.append(metrics.hit_at_k(r, K))
MAP.append(metrics.AP_at_k(r, K, len(user_pos_test)))
return {'recall': np.array(recall), 'precision': np.array(precision), 'ndcg': np.array(ndcg),
'hit_ratio': np.array(hit_ratio), 'MAP': np.array(MAP), 'auc': auc}
preds[u] = list(top_n)
# get actual interaction info. of test users
test_ur = defaultdict(list)
for u in test_set.user.unique():
test_ur[u] = test_set.loc[test_set.user == u,
'item'].values.tolist()
for u in preds.keys():
preds[u] = [1 if e in test_ur[u] else 0 for e in preds[u]]
# calculate metrics
precision_k = np.mean(
[precision_at_k(r, args.topk) for r in preds.values()])
fnl_precision.append(precision_k)
recall_k = np.mean([
recall_at_k(r, len(test_ur[u]), args.topk)
for u, r in preds.items()
])
fnl_recall.append(recall_k)
map_k = map_at_k(list(preds.values()))
fnl_map.append(map_k)
ndcg_k = np.mean([ndcg_at_k(r, args.topk) for r in preds.values()])
fnl_ndcg.append(ndcg_k)
hr_k = hr_at_k(list(preds.values()), list(preds.keys()), test_ur)
fnl_hr.append(hr_k)
mrr_k = mrr_at_k(list(preds.values()))
fnl_mrr.append(mrr_k)
actual_cands = set(candidates[u])
neg_item_pool = set(item_pool) - set(test_ur[u]) - set(ur[u])
neg_cands = random.sample(neg_item_pool, max_i_num - len(candidates[u]))
cands = actual_cands | set(neg_cands)
else:
cands = random.sample(candidates[u], max_i_num)
pred_rates = algo.user_vec[u, :].dot(algo.item_vec).toarray()[0, list(cands)]
rec_idx = np.argsort(pred_rates)[::-1][:args.topk]
preds[u] = list(np.array(list(cands))[rec_idx])
for u in preds.keys():
preds[u] = [1 if i in test_ur[u] else 0 for i in preds[u]]
precision_k = np.mean([precision_at_k(r, args.topk) for r in preds.values()])
fnl_precision.append(precision_k)
recall_k = np.mean([recall_at_k(r, len(test_ur[u]), args.topk) for u, r in preds.items()])
fnl_recall.append(recall_k)
map_k = map_at_k(list(preds.values()))
fnl_map.append(map_k)
ndcg_k = np.mean([ndcg_at_k(r, args.topk) for r in preds.values()])
fnl_ndcg.append(ndcg_k)
hr_k = hr_at_k(list(preds.values()), list(preds.keys()), test_ur)
fnl_hr.append(hr_k)
mrr_k = mrr_at_k(list(preds.values()))
fnl_mrr.append(mrr_k)
for i in range(len(val_kpi)):
# get top-N list for test users
preds = reco.predict(test_set)
# get actual interaction info. of test users
test_ur = defaultdict(list)
for u in test_set.user.unique():
test_ur[u] = test_set.loc[test_set.user == u,
'item'].values.tolist()
for u in preds.keys():
preds[u] = [1 if e in test_ur[u] else 0 for e in preds[u]]
# calculate metrics
precision_k = np.mean([precision_at_k(r, k) for r in preds.values()])
fnl_precision.append(precision_k)
recall_k = np.mean(
[recall_at_k(r, len(test_ur[u]), k) for u, r in preds.items()])
fnl_recall.append(recall_k)
map_k = map_at_k(list(preds.values()))
fnl_map.append(map_k)
ndcg_k = np.mean([ndcg_at_k(r, k) for r in preds.values()])
fnl_ndcg.append(ndcg_k)
hr_k = hr_at_k(list(preds.values()), list(preds.keys()), test_ur)
fnl_hr.append(hr_k)
mrr_k = mrr_at_k(list(preds.values()))
fnl_mrr.append(mrr_k)
print('---------------------------------')
top_n = np.array(test_u_is[u])[rec_idx]
preds[u] = list(top_n)
# get actual interaction info. of test users
ur = defaultdict(list)
for u in test_set.user.unique():
ur[u] = test_set.loc[test_set.user == u, 'item'].values.tolist()
for u in preds.keys():
preds[u] = [1 if e in ur[u] else 0 for e in preds[u]]
# calculate metrics
precision_k = np.mean(
[precision_at_k(r, args.topk) for r in preds.values()])
fnl_precision.append(precision_k)
recall_k = np.mean(
[recall_at_k(r, len(ur[u]), args.topk) for u, r in preds.items()])
fnl_recall.append(recall_k)
map_k = map_at_k(list(preds.values()))
fnl_map.append(map_k)
ndcg_k = np.mean([ndcg_at_k(r, args.topk) for r in preds.values()])
fnl_ndcg.append(ndcg_k)
hr_k = hr_at_k(list(preds.values()), list(preds.keys()), ur)
fnl_hr.append(hr_k)
mrr_k = mrr_at_k(list(preds.values()))
fnl_mrr.append(mrr_k)
gc.collect()