데모¶

라이브러리 import 및 설정¶

%reload_ext autoreload
%autoreload 2
%matplotlib inline

from matplotlib import pyplot as plt
from matplotlib import rcParams
import numpy as np
from pathlib import Path
import pandas as pd
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score
from sklearn.model_selection import StratifiedKFold
from sklearn.preprocessing import StandardScaler
import seaborn as sns
import warnings

rcParams['figure.figsize'] = (16, 8)
plt.style.use('fivethirtyeight')
pd.set_option('max_columns', 100)
pd.set_option("display.precision", 4)
warnings.simplefilter('ignore')

학습데이터 로드¶

03-pandas-eda.ipynb에서 생성한 feature.csv 피처파일 사용

data_dir = Path('../data/dacon-dku')
feature_dir = Path('../build/feature')
val_dir = Path('../build/val')
tst_dir = Path('../build/tst')
sub_dir = Path('../build/sub')

trn_file = data_dir / 'train.csv'
tst_file = data_dir / 'test.csv'
sample_file = data_dir / 'sample_submission.csv'

target_col = 'class'
n_fold = 5
n_class = 3
seed = 42

model_names = ['lrcv_polyfeature', 'rfcv_feature', 'lgbcv_feature']

algo_name = 'lrcv'
feature_name = 'stacking1'
model_name = f'{algo_name}_{feature_name}'

feature_file = feature_dir / f'{feature_name}.csv'
p_val_file = val_dir / f'{model_name}.val.csv'
p_tst_file = tst_dir / f'{model_name}.tst.csv'
sub_file = sub_dir / f'{model_name}.csv'

Stacking Feature 생성¶

trn = []
tst = []
for model in model_names:
    trn.append(np.loadtxt(val_dir / f'{model}.val.csv', delimiter=','))
    tst.append(np.loadtxt(tst_dir / f'{model}.tst.csv', delimiter=','))
    
trn = np.hstack(trn)
tst = np.hstack(tst)
print(trn.shape, tst.shape)

(320000, 9) (80000, 9)

feature_names = []
for model in model_names:
    feature_names += [f'{model}_class0', f'{model}_class1', f'{model}_class2']
    
feature_names

['lrcv_polyfeature_class0',
 'lrcv_polyfeature_class1',
 'lrcv_polyfeature_class2',
 'rfcv_feature_class0',
 'rfcv_feature_class1',
 'rfcv_feature_class2',
 'lgbcv_feature_class0',
 'lgbcv_feature_class1',
 'lgbcv_feature_class2']

y = pd.read_csv(trn_file, index_col=0, usecols=['id', target_col]).values.flatten()
y.shape

(320000,)

Stratified K-Fold Cross Validation¶

cv = StratifiedKFold(n_splits=n_fold, shuffle=True, random_state=seed)

Logistic Regression 모델 학습¶

p_val = np.zeros((trn.shape[0], n_class))
p_tst = np.zeros((tst.shape[0], n_class))
for i, (i_trn, i_val) in enumerate(cv.split(trn, y), 1):
    print(f'training model for CV #{i}')
    clf = LogisticRegression(multi_class='multinomial')
    clf.fit(trn[i_trn], y[i_trn])
    p_val[i_val, :] = clf.predict_proba(trn[i_val])
    p_tst += clf.predict_proba(tst) / n_fold

training model for CV #1
training model for CV #2
training model for CV #3
training model for CV #4
training model for CV #5

print(f'{accuracy_score(y, np.argmax(p_val, axis=1)) * 100:.4f}%')

93.0947%

print(p_val.shape, p_tst.shape)

(320000, 3) (80000, 3)

np.savetxt(p_val_file, p_val, fmt='%.6f', delimiter=',')
np.savetxt(p_tst_file, p_tst, fmt='%.6f', delimiter=',')

피처 중요도 시각화¶

clf.coef_.shape

(3, 9)

imp = pd.DataFrame({'feature': feature_names, 'importance': clf.coef_[1, :].T})
imp = imp.sort_values('importance').set_index('feature')
imp.plot(kind='barh', figsize=(8, 8))

<matplotlib.axes._subplots.AxesSubplot at 0x7fd4899b7e10>

../_images/08-stacking-lr_23_1.png

제출 파일 생성¶

sub = pd.read_csv(sample_file, index_col=0)
print(sub.shape)
sub.head()

(80000, 1)

	class
id
320000	0
320001	0
320002	0
320003	0
320004	0

sub[target_col] = np.argmax(p_tst, axis=1)
sub.head()

	class
id
320000	2
320001	0
320002	2
320003	0
320004	2

sub[target_col].value_counts()

  41194
  29968
   8838
Name: class, dtype: int64

sub.to_csv(sub_file)

데모 데모