def one_fold():
path = 'D:/output/topk/random/'
# path = '/mnt/sata_disk/hossein/wsdmcup2020/output/'
k_list = [3, 5, 10, 100]
random_sample_size = 1000
top_relevant_paper_size = 100
feature_types = ['bm.1_1_gram', 'bm.1_2_gram']
feature_types += ['tf.1_1_gram', 'tf.1_2_gram']
feature_types += ['tfidf.1_1_gram', 'tfidf.1_2_gram']
feature_types += ['emd.50', 'emd.100', 'emd.200', 'emd.300']
feature_types += [
'glv.emd.50', 'glv.emd.100', 'glv.emd.200', 'glv.emd.300'
]
feature_types += ['d2v.50', 'd2v.100', 'd2v.200', 'd2v.300']
paper_info = ['title', 'abstract', 'journal']
workbook = xlsxwriter.Workbook(
path + 'topk/random/eval_one_fold.{}.{}.xlsx'.format(
random_sample_size, top_relevant_paper_size))
ws_avg_map_k = workbook.add_worksheet('avg_map_k')
ws_var_map_k = workbook.add_worksheet('var_map_k')
ws_avg_suc_k = workbook.add_worksheet('avg_suc_k')
ws_var_suc_k = workbook.add_worksheet('var_suc_k')
row = 0
for info in paper_info:
for feature_type in feature_types:
row += 1
for i, k in enumerate(k_list):
ws_avg_map_k.write(0, i + 1, 'avg_map@{}'.format(k))
ws_avg_map_k.write(row, 0, '{}.{}'.format(info, feature_type))
ws_var_map_k.write(0, i + 1, 'var_map@{}'.format(k))
ws_var_map_k.write(row, 0, '{}.{}'.format(info, feature_type))
ws_avg_suc_k.write(0, i + 1, 'avg_suc@{}'.format(k))
ws_avg_suc_k.write(row, 0, '{}.{}'.format(info, feature_type))
ws_var_suc_k.write(0, i + 1, 'avg_suc@{}'.format(k))
ws_var_suc_k.write(row, 0, '{}.{}'.format(info, feature_type))
predictions = pd.read_csv(
path + 'topk/random/train.cos.{}.{}.{}.top{}'.format(
info, feature_type, random_sample_size,
top_relevant_paper_size),
header=0)
trues = [[t] for t in predictions['paper_id'].tolist()]
preds = [str(p).split(',') for p in predictions['topk']]
mean_suc_k(true=trues, pred=preds, k=k)
print('map@{}:{}.{}:{}'.format(
k, info, feature_type,
mean_ap_k(true=trues, pred=preds, k=k)))
print('suc@{}:{}.{}:{}'.format(
k, info, feature_type,
mean_suc_k(true=trues, pred=preds, k=k)))
ws_avg_map_k.write(row, i + 1,
mean_ap_k(true=trues, pred=preds, k=k))
ws_avg_suc_k.write(row, i + 1,
mean_suc_k(true=trues, pred=preds, k=k))
workbook.close()
def one_fold_mixed():
path = 'D:/output/topk/random/'
# path = '/mnt/sata_disk/hossein/wsdmcup2020/output/'
k_list = [3, 5, 10, 100]
random_sample_size = 1000
top_relevant_paper_size = 100
feature_types = [
'bm.1_1_gram', 'bm.1_2_gram', 'tfidf.1_1_gram', 'tfidf.1_2_gram'
]
paper_info = ['title', 'abstract']
k = 100
workbook = xlsxwriter.Workbook(
path + 'eval_one_fold.mix.{}.{}.xlsx'.format(random_sample_size,
top_relevant_paper_size))
ws_avg_map_k = workbook.add_worksheet('avg_map_k')
ws_var_map_k = workbook.add_worksheet('var_map_k')
ws_avg_suc_k = workbook.add_worksheet('avg_suc_k')
ws_var_suc_k = workbook.add_worksheet('var_suc_k')
row = 0
for info1 in paper_info:
for feature_type1 in feature_types:
for info2 in paper_info:
for feature_type2 in feature_types:
if info1 == info2 and feature_type1 == feature_type2:
continue
logger.info(
'train.cos.{}.{} mixed with train.cos.{}.{} ...'.
format(info1, feature_type1, info2, feature_type2))
row += 1
for i, k in enumerate(k_list):
ws_avg_map_k.write(0, i + 1, 'avg_map@{}'.format(k))
ws_avg_map_k.write(
row, 0,
'{}.{}.mix.{}.{}'.format(info1, feature_type1,
info2, feature_type2))
ws_var_map_k.write(0, i + 1, 'var_map@{}'.format(k))
ws_var_map_k.write(
row, 0,
'{}.{}.mix.{}.{}'.format(info1, feature_type1,
info2, feature_type2))
ws_avg_suc_k.write(0, i + 1, 'avg_suc@{}'.format(k))
ws_avg_suc_k.write(
row, 0,
'{}.{}.mix.{}.{}'.format(info1, feature_type1,
info2, feature_type2))
ws_var_suc_k.write(0, i + 1, 'avg_suc@{}'.format(k))
ws_var_suc_k.write(
row, 0,
'{}.{}.mix.{}.{}'.format(info1, feature_type1,
info2, feature_type2))
predictions = pd.read_csv(
path + 'train.cos.{}.{}.mix.{}.{}.{}.top{}'.format(
info1, feature_type1, info2, feature_type2,
random_sample_size, top_relevant_paper_size))
trues = [[t] for t in predictions['paper_id'].tolist()]
preds = [
str(p).split(',') for p in predictions['topk']
]
mean_suc_k(true=trues, pred=preds, k=k)
ws_avg_map_k.write(
row, i + 1, mean_ap_k(true=trues, pred=preds, k=k))
ws_avg_suc_k.write(
row, i + 1, mean_suc_k(true=trues, pred=preds,
k=k))
workbook.close()