def fit_model(self, data, target, test):
clf = XGBRegressor(learning_rate=self.learning_rate,
n_estimators=self.n_estimators,
max_depth=self.max_depth,
min_child_weight=self.min_child_weight,
gamma=self.gamma,
subsample=self.subsample,
colsample_bytree=self.colsample_bytree,
objective=self.objective,
nthread=self.nthread,
scale_pos_weight=self.scale_pos_weight,
reg_alpha=self.reg_alpha,
reg_lambda=self.reg_lambda,
seed=self.seed)
data = np.array(data).astype(float)
scaler = MinMaxScaler()
temp = scaler.fit(data)
data = scaler.transform(data)
test = scaler.transform(test)
target = scaler.fit_transform(target)
clf.fit(data, target)
new_feature = clf.apply(data)
new_test = clf.apply(test)
X_train_new = self.mergeToOne(pd.DataFrame(data), new_feature)
X_test_new = self.mergeToOne(pd.DataFrame(test), new_test)
X_train_new = pd.DataFrame(X_train_new)
X_test_new = pd.DataFrame(X_test_new)
return X_train_new, target, X_test_new
def fit_model_split(self, X_train, y_train, X_test, y_test):
##X_train_1用于生成模型 X_train_2用于和新特征组成新训练集合
X_train_1, X_train_2, y_train_1, y_train_2 = train_test_split(
X_train, y_train, test_size=0.6, random_state=0)
clf = XGBRegressor(learning_rate=self.learning_rate,
n_estimators=self.n_estimators,
max_depth=self.max_depth,
min_child_weight=self.min_child_weight,
gamma=self.gamma,
subsample=self.subsample,
colsample_bytree=self.colsample_bytree,
objective=self.objective,
nthread=self.nthread,
scale_pos_weight=self.scale_pos_weight,
reg_alpha=self.reg_alpha,
reg_lambda=self.reg_lambda,
seed=self.seed)
clf.fit(X_train_1, y_train_1)
# y_pre = clf.predict(X_train_2)
# y_pro = clf.predict_proba(X_train_2)[:, 1]
# print
# "pred_leaf=T AUC Score : %f" % metrics.roc_auc_score(y_train_2, y_pro)
# print
# "pred_leaf=T Accuracy : %.4g" % metrics.accuracy_score(y_train_2, y_pre)
new_feature = clf.apply(X_train_2)
X_train_new2 = self.mergeToOne(X_train_2, new_feature)
new_feature_test = clf.apply(X_test)
X_test_new = self.mergeToOne(X_test, new_feature_test)
print
"Training set of sample size 0.4 fewer than before"
return X_train_new2, y_train_2, X_test_new, y_test
def fit_model(self, X_train, y_train, X_test, y_test):
clf = XGBRegressor(learning_rate=self.learning_rate,
n_estimators=self.n_estimators,
max_depth=self.max_depth,
min_child_weight=self.min_child_weight,
gamma=self.gamma,
subsample=self.subsample,
colsample_bytree=self.colsample_bytree,
objective=self.objective,
nthread=self.nthread,
scale_pos_weight=self.scale_pos_weight,
reg_alpha=self.reg_alpha,
reg_lambda=self.reg_lambda,
seed=self.seed)
clf.fit(X_train, y_train)
# y_pre = clf.predict(X_test)
# y_pro = clf.predict_proba(X_test)[:, 1]
# print
# "pred_leaf=T AUC Score : %f" % metrics.roc_auc_score(y_test, y_pro)
# print("pred_leaf=T Accuracy : %.4g" % metrics.accuracy_score(y_test, y_pre))
new_feature = clf.apply(X_train)
X_train_new = self.mergeToOne(X_train, new_feature)
new_feature_test = clf.apply(X_test)
X_test_new = self.mergeToOne(X_test, new_feature_test)
print
"Training set sample number remains the same"
return X_train_new, y_train, X_test_new, y_test
def fit_model_split(self,X_train,y_train,X_test,y_test):
##X_train_1用于生成模型 X_train_2用于和新特征组成新训练集合
X_train_1, X_train_2, y_train_1, y_train_2 = train_test_split(X_train, y_train, test_size=0.6, random_state=0)
clf = XGBRegressor(
learning_rate =self.learning_rate,
n_estimators=self.n_estimators,
max_depth=self.max_depth,
min_child_weight=self.min_child_weight,
gamma=self.gamma,
subsample=self.subsample,
colsample_bytree=self.colsample_bytree,
objective= self.objective,
nthread=self.nthread,
scale_pos_weight=self.scale_pos_weight,
reg_alpha=self.reg_alpha,
reg_lambda=self.reg_lambda,
seed=self.seed)
clf.fit(X_train_1, y_train_1)
new_feature= clf.apply(X_train_2)
X_train_new2=self.mergeToOne(X_train_2,new_feature)
new_feature_test= clf.apply(X_test)
X_test_new=self.mergeToOne(X_test,new_feature_test)
return X_train_new2,y_train_2,X_test_new,y_test
def get_XgbRegressor(train_data,
train_target,
test_data,
feature_names,
parameters,
early_stopping_rounds,
num_folds,
eval_metric,
model_name='model',
stratified=False):
'''
:param train_data: 一定是numpy
:param train_target:
:param parameters:
:param round:
:param k:
:param eval_metrics:自定义 or 内置字符串
:return:
'''
reg = XGBRegressor()
reg.set_params(**parameters)
# 定义一些变量
oof_preds = np.zeros((train_data.shape[0], ))
sub_preds = np.zeros((test_data.shape[0], ))
feature_importance_df = pd.DataFrame()
cv_result = []
# K-flod
if stratified:
folds = StratifiedKFold(n_splits=num_folds,
shuffle=True,
random_state=1234)
else:
folds = KFold(n_splits=num_folds, shuffle=True, random_state=1234)
X_train_newfeature = np.zeros((1, 1))
for n_flod, (train_index,
val_index) in enumerate(folds.split(train_data,
train_target)):
train_X = train_data[train_index]
val_X = train_data[val_index]
train_Y = train_target[train_index]
val_Y = train_target[val_index]
# 参数初步定之后划分20%为验证集，准备一个watchlist 给train和validation set ,设置num_round 足够大（比如100000），以至于你能发现每一个round 的验证集预测结果，
# 如果在某一个round后 validation set 的预测误差上升了，你就可以停止掉正在运行的程序了。
watchlist = [(train_X, train_Y), (val_X, val_Y)]
# early_stop 看validate的eval是否下降，这时候必须传eval_set,并取eval_set的最后一个作为validate
reg.fit(train_X,
train_Y,
early_stopping_rounds=early_stopping_rounds,
eval_set=watchlist,
eval_metric=eval_metric)
## 生成gbdt新特征
new_feature = reg.apply(val_X)
if X_train_newfeature.shape[0] == 1:
X_train_newfeature = mergeToOne(val_X, new_feature)
else:
X_train_newfeature = mergeToOne(val_X, new_feature)
X_train_newfeature = np.concatenate(
(X_train_newfeature, mergeToOne(new_feature, val_X)), axis=0)
print(X_train_newfeature)
# 获得每次的预测值补充
oof_preds[val_index] = reg.predict(val_X)
# 获得预测的平均值，这里直接加完再除m
sub_preds += reg.predict(test_data)
result = mean_absolute_error(val_Y, reg.predict(val_X))
print('Fold %2d macro-f1 : %.6f' % (n_flod + 1, result))
cv_result.append(round(result, 5))
gc.collect()
# 默认就是gain 如果要修改要再参数定义中修改importance_type
# 保存特征重要度
gain = reg.feature_importances_
fold_importance_df = pd.DataFrame({
'feature': feature_names,
'gain': 100 * gain / gain.sum(),
'fold': n_flod,
}).sort_values('gain', ascending=False)
feature_importance_df = pd.concat(
[feature_importance_df, fold_importance_df], axis=0)
# 进行保存
sub_preds = sub_preds / folds.n_splits
new_feature = reg.apply(test_data)
X_test_newfeature = mergeToOne(test_data, new_feature)
if not os.path.isdir('./sub'):
os.makedirs('./sub')
pd.DataFrame(oof_preds,
columns=['class'
]).to_csv('./sub/val_{}.csv'.format(model_name),
index=False)
pd.DataFrame(sub_preds,
columns=['class'
]).to_csv('./sub/test_{}.csv'.format(model_name),
index=False)
print('cv_result', cv_result)
if not os.path.isdir('./gbdt_newfeature'):
os.makedirs('./gbdt_newfeature')
np.save("./gbdt_newfeature/train_newfeature.npy", X_train_newfeature)
np.save("./gbdt_newfeature/test_newfeature.npy", X_test_newfeature)
save_importances(feature_importance_df, model_name)
return reg, sub_preds