-
Notifications
You must be signed in to change notification settings - Fork 0
/
local_blends.py
34 lines (28 loc) · 1.25 KB
/
local_blends.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
import pandas as pd
from sklearn.preprocessing import minmax_scale
labels = ["toxic", "severe_toxic", "obscene", "threat", "insult", "identity_hate"]
df_9836 = pd.read_csv('sub_h_n_consolidated_filtered_inter.csv')
df_9830 = pd.read_csv('gru_attn_300p_300d_lc_submission.csv')
df_9782 = pd.read_csv('submission-lr-no-cw.csv')
df_svm = pd.read_csv('submission-lsvm-no-cw.csv')
df_han_cnn = pd.read_csv('han_class_cnn.csv')
for label in labels:
df_9836[label] = minmax_scale(df_9836[label])
df_9830[label] = minmax_scale(df_9830[label])
df_9782[label] = minmax_scale(df_9782[label])
df_svm[label] = minmax_scale(df_svm[label])
df_han_cnn[label] = minmax_scale(df_han_cnn[label])
submission = pd.DataFrame()
submission['id'] = df_9836['id']
"""
submission[labels] = (df_9836[labels]*9 + \
df_9830[labels]*9 + \
df_9782[labels]*2) / 20
submission[labels] = (df_9836[labels]*8 + \
df_9830[labels]*7 + \
df_9782[labels]*5) / 20
"""
submission[labels] = (df_9836[labels]*6 +df_9830[labels]*6 + df_han_cnn[labels]*6) / 18
#for label in labels:
# submission[label] = minmax_scale(submission[label])
submission.to_csv('tc_blend_weighted_lrhigh_hancnn.csv', index=False)