In [1]:

!pip install -q iterative-stratification

WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv

In [52]:

import pandas as pd
from sklearn.model_selection import KFold
from iterstrat.ml_stratifiers import MultilabelStratifiedKFold

df = pd.read_csv('/root/.cache/data/train.csv')

In [53]:

dfx = pd.get_dummies(df, columns=["discourse_type"]).head(5)
dfx

Out[53]:

	id	discourse_id	discourse_start	discourse_end	discourse_text	discourse_type_num	predictionstring	discourse_type_Claim	discourse_type_Evidence	discourse_type_Lead	discourse_type_Position
0	423A1CA112E2	1.622628e+12	8.0	229.0	Modern humans today are always on their phone....	Lead 1	1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1...	0	0	1	0
1	423A1CA112E2	1.622628e+12	230.0	312.0	They are some really bad consequences when stu...	Position 1	45 46 47 48 49 50 51 52 53 54 55 56 57 58 59	0	0	0	1
2	423A1CA112E2	1.622628e+12	313.0	401.0	Some certain areas in the United States ban ph...	Evidence 1	60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75	0	1	0	0
3	423A1CA112E2	1.622628e+12	402.0	758.0	When people have phones, they know about certa...	Evidence 2	76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 9...	0	1	0	0
4	423A1CA112E2	1.622628e+12	759.0	886.0	Driving is one of the way how to get around. P...	Claim 1	139 140 141 142 143 144 145 146 147 148 149 15...	1	0	0	0

In [54]:

dfx = pd.get_dummies(df, columns=["discourse_type"]).groupby(["id"], as_index=False).sum()
dfx

Out[54]:

	id	discourse_id	discourse_start	discourse_end	discourse_type_Claim	discourse_type_Concluding Statement	discourse_type_Counterclaim	discourse_type_Evidence	discourse_type_Lead	discourse_type_Position	discourse_type_Rebuttal
0	0000D23A521A	1.294188e+13	4166.0	5506.0	1	1	1	3	0	1	1
1	00066EA9880D	1.458994e+13	12618.0	16058.0	3	1	0	3	1	1	0
2	000E6DE9E817	1.940756e+13	8760.0	10092.0	5	1	1	3	0	1	1
3	001552828BD0	1.622844e+13	12881.0	15580.0	4	0	0	4	1	1	0
4	0016926B079C	1.783190e+13	5102.0	6414.0	7	0	0	3	0	1	0
...	...	...	...	...	...	...	...	...	...	...	...
15589	FFF1442D6698	1.618644e+13	14374.0	17948.0	2	1	1	3	1	1	1
15590	FFF1ED4F8544	1.454313e+13	6944.0	9435.0	5	0	0	2	1	1	0
15591	FFF868E06176	1.456920e+13	8210.0	10507.0	3	1	0	3	1	1	0
15592	FFFD0AF13501	1.295859e+13	4408.0	5395.0	4	1	0	2	0	1	0
15593	FFFF80B8CC2F	1.617042e+12	0.0	990.0	0	0	0	1	0	0	0

15594 rows × 11 columns

In [55]:

dfx.columns

Out[55]:

Index(['id', 'discourse_id', 'discourse_start', 'discourse_end',
       'discourse_type_Claim', 'discourse_type_Concluding Statement',
       'discourse_type_Counterclaim', 'discourse_type_Evidence',
       'discourse_type_Lead', 'discourse_type_Position',
       'discourse_type_Rebuttal'],
      dtype='object')

In [56]:

cols = [c for c in dfx.columns if c.startswith("discourse_type") or c =="id" and c !=  "discourse_type_num"]
cols

Out[56]:

['id',
 'discourse_type_Claim',
 'discourse_type_Concluding Statement',
 'discourse_type_Counterclaim',
 'discourse_type_Evidence',
 'discourse_type_Lead',
 'discourse_type_Position',
 'discourse_type_Rebuttal']

In [57]:

dfx = dfx[cols]
dfx

Out[57]:

	id	discourse_type_Claim	discourse_type_Concluding Statement	discourse_type_Counterclaim	discourse_type_Evidence	discourse_type_Lead	discourse_type_Position	discourse_type_Rebuttal
0	0000D23A521A	1	1	1	3	0	1	1
1	00066EA9880D	3	1	0	3	1	1	0
2	000E6DE9E817	5	1	1	3	0	1	1
3	001552828BD0	4	0	0	4	1	1	0
4	0016926B079C	7	0	0	3	0	1	0
...	...	...	...	...	...	...	...	...
15589	FFF1442D6698	2	1	1	3	1	1	1
15590	FFF1ED4F8544	5	0	0	2	1	1	0
15591	FFF868E06176	3	1	0	3	1	1	0
15592	FFFD0AF13501	4	1	0	2	0	1	0
15593	FFFF80B8CC2F	0	0	0	1	0	0	0

15594 rows × 8 columns

In [58]:

mskf = MultilabelStratifiedKFold(n_splits=5, shuffle=True, random_state=42)
labels = [c for c in dfx.columns if c != "id"]
dfx_labels = dfx[labels]
dfx_labels

Out[58]:

	discourse_type_Claim	discourse_type_Concluding Statement	discourse_type_Counterclaim	discourse_type_Evidence	discourse_type_Lead	discourse_type_Position	discourse_type_Rebuttal
0	1	1	1	3	0	1	1
1	3	1	0	3	1	1	0
2	5	1	1	3	0	1	1
3	4	0	0	4	1	1	0
4	7	0	0	3	0	1	0
...	...	...	...	...	...	...	...
15589	2	1	1	3	1	1	1
15590	5	0	0	2	1	1	0
15591	3	1	0	3	1	1	0
15592	4	1	0	2	0	1	0
15593	0	0	0	1	0	0	0

15594 rows × 7 columns

In [59]:

dfx["kfold"] = -1

/tmp/ipykernel_27806/3539168384.py:1: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  dfx["kfold"] = -1

In [86]:

df = pd.read_csv('/root/.cache/data/train.csv')

dfx = pd.get_dummies(df, columns=["discourse_type"]).groupby(["id"], as_index=False).sum()
cols = [c for c in dfx.columns if c.startswith("discourse_type_") or c == "id" and c != "discourse_type_num"]
dfx = dfx[cols]

mskf = MultilabelStratifiedKFold(n_splits=10, shuffle=True, random_state=42)
labels = [c for c in dfx.columns if c != "id"]
dfx_labels = dfx[labels]
dfx["kfold"] = -1
dfx

Out[86]:

	id	discourse_type_Claim	discourse_type_Concluding Statement	discourse_type_Counterclaim	discourse_type_Evidence	discourse_type_Lead	discourse_type_Position	discourse_type_Rebuttal	kfold
0	0000D23A521A	1	1	1	3	0	1	1	-1
1	00066EA9880D	3	1	0	3	1	1	0	-1
2	000E6DE9E817	5	1	1	3	0	1	1	-1
3	001552828BD0	4	0	0	4	1	1	0	-1
4	0016926B079C	7	0	0	3	0	1	0	-1
...	...	...	...	...	...	...	...	...	...
15589	FFF1442D6698	2	1	1	3	1	1	1	-1
15590	FFF1ED4F8544	5	0	0	2	1	1	0	-1
15591	FFF868E06176	3	1	0	3	1	1	0	-1
15592	FFFD0AF13501	4	1	0	2	0	1	0	-1
15593	FFFF80B8CC2F	0	0	0	1	0	0	0	-1

15594 rows × 9 columns

In [87]:

for fold, (trn_, val_) in enumerate(mskf.split(dfx, dfx_labels)):
    print(len(trn_), len(val_))
    dfx.loc[val_, "kfold"] = fold

df = df.merge(dfx[["id", "kfold"]], on="id", how="left")
print(df.kfold.value_counts())
# df.to_csv("train_folds.csv", index=False)

14036 1558
14036 1558
14033 1561
14035 1559
14031 1563
14035 1559
14034 1560
14036 1558
14036 1558
14034 1560
6    14633
9    14532
7    14529
8    14509
5    14466
3    14431
4    14365
1    14358
2    14271
0    14199
Name: kfold, dtype: int64

In [88]:

df.groupby(["kfold"]).count()

Out[88]:

	id	discourse_id	discourse_start	discourse_end	discourse_text	discourse_type	discourse_type_num	predictionstring
kfold
0	14199	14199	14199	14199	14199	14199	14199	14199
1	14358	14358	14358	14358	14358	14358	14358	14358
2	14271	14271	14271	14271	14271	14271	14271	14271
3	14431	14431	14431	14431	14431	14431	14431	14431
4	14365	14365	14365	14365	14365	14365	14365	14365
5	14466	14466	14466	14466	14466	14466	14466	14466
6	14633	14633	14633	14633	14633	14633	14633	14633
7	14529	14529	14529	14529	14529	14529	14529	14529
8	14509	14509	14509	14509	14509	14509	14509	14509
9	14532	14532	14532	14532	14532	14532	14532	14532

In [ ]:

Creating folds properly

Comments