'n_estimators': 500,
'max_depth': 3,
'random_state': SEED
}
tuned_params = {
'n_estimators': [400, 600, 800, 1000, 1200],
'max_depth': [3, 4, 5, 6, 7],
'random_state': [1, 2, 3, 4, 5]
}
X_train, y_train, X_test, test_id, feature_names = load_data(processedDataPath)
X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, random_state=SEED, test_size=TEST_SIZE)
trainer = ortTrainer(modelClass=RandomForestClassifier,
params=other_params,
tuned_params=tuned_params,
isupdate=True, istune=True,
modelName='rf', cv=5)
trainer.fit(X_train, y_train)
output = trainer.getOutput(X_test, test_id, X_val, y_val)
bestResult = pd.read_csv('currentBest.txt', header=None, sep='\t')
bestResult.columns = ['id', 'prob']
pccs = pearsonr(output['prob'], bestResult['prob'])
print("Score prediction: {}".format(pccs))
print(trainer.bestParams)
sorted_importances, sorted_featurenames = zip(
*sorted(zip(trainer.model.feature_importances_, feature_names), reverse=True)
tuned_params = {
"learning_rate": [0.02, 0.2, 0.5, 0.9], # shrinkage
"max_depth": [8, 12, 15, 20],
"subsample": [0.6, 0.7, 0.8, 0.9],
"colsample_bytree": [0.3, 0.5, 0.7, 0.9],
}
X_train, y_train, X_test, test_id, feature_names = load_data(processedDataPath)
X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, random_state=SEED, test_size=TEST_SIZE)
trainer = ortTrainer(modelClass=XGBClassifier ,
params=params,
tuned_params=tuned_params,
isupdate=True,
istune=True,
modelName='xgb-ordinal')
trainer.fit( X_train, y_train,
eval_set=[(X_val, y_val)],
early_stopping_rounds=100,
eval_metric="auc",
verbose=True
)
trainer.getOutput(X_test, test_id, X_val, y_val)
modelName='adb', cv=5)
trainers = []
adbtrainer = TrainerWrapper(baseTrainer, adbtrainer_params)
trainers.append(adbtrainer)
#extratrainer= TrainerWrapper(baseTrainer, extratrainer_params)
rftrainer = TrainerWrapper(ortTrainer, rftrainer_params)
trainers.append(rftrainer)
lgbtrainer = XgBoostWrapper(ortTrainer, lgbtrainer_params)
trainers.append(lgbtrainer)
features = [get_oof(trainer, X_train, y_train, X_test) for trainer in trainers]
X_train_new = np.hstack(feature[0] for feature in features)
X_test_new = np.hstack(feature[1] for feature in features)
finaltrainer = ortTrainer(modelClass=LGBMClassifier,
params=lgb_params, tuned_params=lgb_tune_params,
isupdate=True, istune=True,
modelName='final', cv=5)
finaltrainer.fit(X_train_new, y_train)
_, X_val_new, _, y_val_new = train_test_split(X_train_new, y_train, random_state=SEED)
finaltrainer.getOutput(X_test_new, test_id, X_val_new, y_val_new)
print("haha")
from baseModel import ortTrainer
from path import *
from utils import load_data
from sklearn.linear_model import LogisticRegression
other_params = {
'n_jobs': 1,
'random_state': 1,
'C': 0.3
}
tuned_params = {
'C': [0.01, 0.05, 0.1, 0.3, 0.5, 0.7]
}
X_train, y_train, X_test, test_id, feature_names = load_data(processedDataPath)
trainer = ortTrainer(modelClass=LogisticRegression,
params=other_params,
tuned_params=tuned_params,
isupdate=True,
istune=True,
modelName='lr')
trainer.fit(X_train, y_train)
trainer.getOutput()
}
tuned_params = {
"learning_rate": [0.02, 0.2, 0.5, 0.8], # shrinkage
"max_depth": [8, 12, 15, 20],
"subsample": [0.5, 0.6, 0.7, 0.8],
'reg_lambda': [0.01, 0.2, 0.5, 0.8],
"colsample_bytree": [0.3, 0.5, 0.7, 0.8],
}
# 训练模型
X_train, y_train, X_test, test_id, feature_names = load_data(reinbalancedDataPath)
X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, random_state=SEED, test_size=TEST_SIZE)
trainer = ortTrainer(modelClass=LGBMClassifier,
params=params,
tuned_params=tuned_params,
isupdate=True,
istune=False,
modelName='lgb-ordinal')
trainer.fit( X_train, y_train,
eval_set=[(X_val, y_val)],
early_stopping_rounds=100,
eval_metric="auc",
verbose=True
)
trainer.getOutput(X_test, test_id, X_val, y_val)
print(trainer.bestParams)
dict(zip(trainer.model.feature_importances_, feature_names))