# -*- coding: utf-8 -*-
import operator
import pandas as pd
import xgboost as xgb
from common.data_fun import get_train_X, get_train_Y
if __name__ == '__main__':
Y = get_train_Y()
xg_train = xgb.DMatrix(get_train_X(), label=Y)
params = {
# 'min_child_weight': 100,
# 'eta': 0.02,
# 'colsample_bytree': 0.7,
# 'max_depth': 7,
# 'subsample': 0.7,
# 'alpha': 1,
# 'gamma': 1,
# 'silent': 1,
# 'verbose_eval': True,
# 'seed': 12,
# 'n_estimators': 100
'max_depth': 7,
'n_estimators': 100
}
rounds = 10
bst = xgb.train(params, xg_train, num_boost_round=rounds)
def lightBGM_model(X, Y):
model = LGBMRegressor(num_leaves=36,
n_estimators=100,
learning_rate=0.07,
random_state=0)
model.fit(X, Y, verbose=True)
return model
if __name__ == '__main__':
best, worst, avg, rounds = 1, 0, 0, 10
for i in range(rounds):
print('round : ' + str(i))
error = lightBGM_model_with_test(get_train_X(), get_train_Y())
print('error : ' + str(error))
if error < best:
best = error
if error > worst:
worst = error
avg += error
avg /= rounds
print('=========================')
print('best > ' + str(best))
print('worst > ' + str(worst))
print('average > ' + str(avg))
'colsample_bytree': 0.6,
'subsample': 0.9
}
xg_train = xgb.DMatrix(train_X, label=train_Y)
xgboost_model = xgb.train(params, xg_train)
xg_test = xgb.DMatrix(test_X)
xg_res = xgboost_model.predict(xg_test)
return xg_res
def stacking(train_X, train_Y, test_X):
train_X1, test_X1 = stacking_layer_1(train_X, train_Y, test_X)
res = stacking_layer_2(train_X1, train_Y, test_X1)
return res
if __name__ == '__main__':
from common.data_fun import get_train_X, get_train_Y
from sklearn.model_selection import train_test_split
x1, x2, y1, y2 = train_test_split(get_train_X(),
get_train_Y(),
test_size=0.2)
predict = stacking(x1, y1, x2)
error = error_fun(predict, y2)[1]
print('stacking error: ' + str(error))
from common.data_fun import get_train_X, get_train_Y
model_list = []
model_list.append({'n_estimators' : 150, 'max_depth' : 6, 'min_child_weight' : 4, 'colsample_bytree' : 0.7, 'subsample' : 0.7})
model_list.append({'n_estimators' : 180, 'max_depth' : 5, 'min_child_weight' : 3, 'colsample_bytree' : 1, 'subsample' : 0.9})
model_list.append({'n_estimators' : 240, 'max_depth' : 6, 'min_child_weight' : 1, 'colsample_bytree' : 0.6, 'subsample' : 0.9})
model_list.append({'n_estimators' : 240, 'max_depth' : 6, 'min_child_weight' : 1, 'colsample_bytree' : 0.6, 'subsample' : 0.9})
model_list.append({'n_estimators' : 200, 'max_depth' : 9, 'min_child_weight' : 2, 'colsample_bytree' : 0.9, 'subsample' : 0.8})
total = 0
cnt = 0
Y = get_train_Y()
# 划分数据集
train_x, test_x, train_y, test_y = train_test_split(get_train_X(), Y, test_size=0.2)
# 训练参数设置和执行
params = model_list[cnt]
cnt += 1
rounds = 10
# 训练
xg_train = xgb.DMatrix(train_x, label = train_y)
xgboost_model = xgb.train(params, xg_train)
# 预测
xg_test = xgb.DMatrix(test_x)
xg_res = xgboost_model.predict(xg_test)
useful_feature = get_useful_features_byLightBGM(X, Y)
X_U = X[useful_feature]
x1, x2, y1, y2 = train_test_split(X_U, Y, test_size=0.2)
y1_log = np.log1p(y1)
model.fit(x1, y1_log)
predict_log = model.predict(x2)
predict = np.expm1(predict_log)
error = error_fun(predict, y2)[1]
return model, error
if __name__ == '__main__':
train_x_full = get_train_X().fillna(get_train_X().median())
train_y = get_train_Y()
best, worst, avg, rounds = 1, 0, 0, 10
for i in range(rounds):
print('round : ' + str(i))
error = linearModel(train_x_full, train_y)
print('error : ' + str(error))
if error < best:
best = error
if error > worst:
worst = error
avg += error
avg /= rounds