# print(f'New ranking: {all_new}')
# print(f'Correct items: {all_correct[f]}')
# print()
# Calculating all metrics
if len(all_correct) > 0:
# print_rerank_count(all_old, all_new)
# all_old, all_new, all_correct[f] = filter_sames(all_old, all_new, all_correct[f])
# print_rerank_count(all_old, all_new)
# if metrics_positive_only:
# all_correct[f], all_old, all_new = keep_pos_only(all_correct[f], all_old, all_new)
if metric_to_use == 'MAP':
for l in lambdas:
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[0]}_initial'].append(
mapk(all_old, all_correct[f], metrics_sizes[0]))
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[0]}_rerank'].append(
mapk(all_new[l], all_correct[f], metrics_sizes[0]))
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[1]}_initial'].append(
mapk(all_old, all_correct[f], metrics_sizes[1]))
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[1]}_rerank'].append(
mapk(all_new[l], all_correct[f], metrics_sizes[1]))
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[2]}_initial'].append(
mapk(all_old, all_correct[f], test_item_amount))
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[2]}_rerank'].append(
mapk(all_new[l], all_correct[f], test_item_amount))
elif metric_to_use == 'NDCG':
for l in lambdas:
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[0]}_initial'].append(
ndcg_at(all_old, all_correct[f], metrics_sizes[0]))
measures_lambda[l][f'{metric_to_use}_{metrics_sizes[0]}_rerank'].append(
for i, r in enumerate(initial_pred_list[f]):
save_correct_items(f, r)
re_rank(extract_predictions(r), fold_item_init_dist[f][i],
rerank_size)
print(
f'{input_rec_path}-{len(all_old)}-{len(all_new[1])}-{len(all_correct)}'
)
# Calculating all metrics
if len(all_correct) > 0:
if metric_to_use == 'MAP':
for l in lambdas:
measures_lambda[l][
f'{metric_to_use}_{metrics_sizes[0]}_initial'].append(
mapk(all_old, all_correct[f],
metrics_sizes[0]))
measures_lambda[l][
f'{metric_to_use}_{metrics_sizes[0]}_rerank'].append(
mapk(all_new[l], all_correct[f],
metrics_sizes[0]))
measures_lambda[l][
f'{metric_to_use}_{metrics_sizes[1]}_initial'].append(
mapk(all_old, all_correct[f],
metrics_sizes[1]))
measures_lambda[l][
f'{metric_to_use}_{metrics_sizes[1]}_rerank'].append(
mapk(all_new[l], all_correct[f],
metrics_sizes[1]))
measures_lambda[l][
f'{metric_to_use}_{metrics_sizes[2]}_initial'].append(
mapk(all_old, all_correct[f], rerank_size))
correct_items = []
for xx in initial_pred['correctItems']:
correct_items.append([int(x) for x in xx.split(';')])
predictions = []
for index, row in initial_pred.iterrows():
cntr = 1
pred = []
while cntr <= test_item_amount:
if type(row[f'p{cntr}']) is str:
pred.append(int(row[f'p{cntr}'].split(';')[0]))
cntr = cntr + 1
predictions.append(pred)
map_10.append(mapk(predictions, correct_items, 10))
map_25.append(mapk(predictions, correct_items, 25))
prec_10.append(precision_at(predictions, correct_items, 10))
prec_25.append(precision_at(predictions, correct_items, 25))
ndcg_10.append(ndcg_at(predictions, correct_items, 10))
ndcg_25.append(ndcg_at(predictions, correct_items, 25))
final_df['test_amount'].append(test_item_amount)
final_df['MAP_10'].append(mean(map_10))
final_df['MAP_25'].append(mean(map_25))
final_df['Prec_10'].append(mean(prec_10))
final_df['Prec_25'].append(mean(prec_25))
final_df['NDCG_10'].append(mean(ndcg_10))
final_df['NDCG_25'].append(mean(ndcg_25))
pd.DataFrame.from_dict(final_df).to_csv(