【データ解析初心者におすすめなSignate練習問題】タイタニック課題をLightGBMで解いた例

conda install -c conda-forge lightgbm

import numpy as np
import pandas as pd
# LightGBM
import lightgbm as lgb

train = pd.read_table('./train.tsv')
test = pd.read_table('./test.tsv')

train.embarked = train.embarked.replace(['C', 'S', 'Q'], [0, 1, 2])
test.embarked = test.embarked.replace(['C', 'S', 'Q'], [0, 1, 2])

import sklearn.preprocessing as sp

X_train=train.drop('id',axis=1)
X_test=test.drop('id',axis=1)
X_train['family']=X_train['sibsp']+X_train['parch']+1
X_test['family']=X_test['sibsp']+X_test['parch']+1
X_train=X_train.drop('sibsp',axis=1)
X_test=X_test.drop('sibsp',axis=1)
X_train=X_train.drop('parch',axis=1)
X_test=X_test.drop('parch',axis=1)

le = sp.LabelEncoder()
le.fit(X_train.sex.unique())
X_train.sex = le.fit_transform(X_train.sex)
le.fit(X_test.sex.unique())
X_test.sex = le.fit_transform(X_test.sex)

# sns.heatmap(X_train.corr(),annot=True)

X_train=X_train.drop('survived',axis=1)

y_train=train['survived']

categorical_features = ['pclass', 'sex','family','fare']

params = {
    'objective': 'binary',
    'max_bin': 30,
    'learning_rate': 0.001,
    'num_leaves': 40
}


# from sklearn.model_selection import KFold
from sklearn.model_selection import StratifiedKFold

y_preds = []
models = []
oof_train = np.zeros((len(X_train),))
cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=1)
for fold_id, (train_index, valid_index) in enumerate(cv.split(X_train, y_train)):
    X_tr = X_train.loc[train_index, :]
    X_val = X_train.loc[valid_index, :]
    y_tr = y_train[train_index]
    y_val = y_train[valid_index]

    lgb_train = lgb.Dataset(X_tr, y_tr,
                            categorical_feature=categorical_features)
    lgb_eval = lgb.Dataset(X_val, y_val,
                           reference=lgb_train,
                           categorical_feature=categorical_features)

    model = lgb.train(params, lgb_train,
                      valid_sets=[lgb_train,lgb_eval],
                      verbose_eval=8000,
                      num_boost_round=5000,
                      early_stopping_rounds=5000)

    oof_train[valid_index] = model.predict(X_val, num_iteration=model.best_iteration)
    y_pred = model.predict(X_test, num_iteration=model.best_iteration)

    y_preds.append(y_pred)
    models.append(model)
    
scores = [
    m.best_score['valid_1']['binary_logloss'] for m in models
]
score = sum(scores) / len(scores)
print('===CV scores===')
print(scores)
print(score)

Training until validation scores don't improve for 5000 rounds
Did not meet early stopping. Best iteration is:
[5000]	training's binary_logloss: 0.34012	valid_1's binary_logloss: 0.370792
Training until validation scores don't improve for 5000 rounds
Did not meet early stopping. Best iteration is:
[5000]	training's binary_logloss: 0.331996	valid_1's binary_logloss: 0.430239
Training until validation scores don't improve for 5000 rounds
Did not meet early stopping. Best iteration is:
[5000]	training's binary_logloss: 0.319917	valid_1's binary_logloss: 0.414441
Training until validation scores don't improve for 5000 rounds
Did not meet early stopping. Best iteration is:
[5000]	training's binary_logloss: 0.329038	valid_1's binary_logloss: 0.394071
Training until validation scores don't improve for 5000 rounds
Did not meet early stopping. Best iteration is:
[5000]	training's binary_logloss: 0.28245	valid_1's binary_logloss: 0.590678
===CV scores===
[0.37079179193015643, 0.4302394765776665, 0.4144412398816655, 0.3940709167584216, 0.5906779783883791]
0.44004428070725776

C_train=train
C_train['count']=1
sum(train[train.survived==1]['count'])/sum(train['count'])

for i in range(446):
    if y_pred[i]>=0.4:
        y_pred[i]=1
    else:
        y_pred[i]=0

import csv
test_data = test.values
with open('./gender_submission.csv', "w") as f:
    writer = csv.writer(f, lineterminator='\n')
    for pid, survived in zip(test_data[:,0].astype(int), y_pred.astype(int)):
        writer.writerow([pid,survived])

lgb.plot_importance(model, height=0.5, figsize=(8,8))