def get_model_feature_importance(model,
model_name="default",
outputpath="./",
importance_type='gain',
num_feature=20):
'''
lightgbm 画importace函数
:param model:
:param model_name:
:param outputpath:
:param importance_type:
:param num_feature:
:return:
'''
try:
outputpath = outputpath + model_name + "_feature_importance.png"
ax = lgb.plot_importance(model,
figsize=(20, 13),
importance_type=importance_type,
title="gain_importance",
max_num_features=num_feature)
plt.savefig(outputpath)
except:
logger.error("create model feature fail.")
return False
else:
logger.info("create model feature sucess.")
return True
def get_model_tree_visual(model,
model_name="default",
tree_index=1,
outputpath="./"):
'''
:param model:
:param model_name:
:param outputpath:
:param importance_type:
:param num_feature:
:return:
'''
try:
outputpath = outputpath + model_name + "_tree.png"
ax = lgb.plot_tree(
model,
tree_index=tree_index,
figsize=(20, 13),
)
plt.savefig(outputpath)
except:
logger.error("create model tree fail.")
return False
else:
logger.info("create model tree sucess.")
return True
def load_model_from_text(model_file):
'''
加载模型
:param model_file:
:return:
'''
try:
bst = lgb.Booster(model_file=model_file)
except:
logger.error("load model fail.")
return False
else:
logger.info("load model from text sucess")
return bst
def save_model_as_text(model, model_name="default", outputpath="./"):
'''
保存模型
:param model:
:param model_name:
:param outputpath:
:return:
'''
try:
outputpath = outputpath + model_name + "_lightgbm.txt"
model.save_model(outputpath)
except:
logger.error("save lightgbm model fail.")
return False
else:
logger.info("save lightgbm model sucess")
return True
def get_model_tree_digraph(model, model_name="default", outputpath="./"):
'''
画出tree 的树结构
:param model:
:param model_name:
:param outputpath:
:param importance_type:
:param num_feature:
:return:
'''
try:
outputpath = outputpath + model_name + "_tree_digraph.gv"
graph = lgb.create_tree_digraph(model, name=model_name)
graph.render(filename=outputpath)
except:
logger.error("create model tree_digrap fail.")
return False
else:
logger.info("create model tree_digrap sucess.")
return True
def train_lightgbm_model(X_train,
Y_train,
feature_name,
model_param,
test_size=0.2):
'''
训练lightgbm 给定参数x,y,x所属特征,分裂的结果,model训练的参数
:param X_train:
:param Y_train:
:param feature_name:
:param test_size:
:param model_param:
:return:
'''
try:
x_train, x_test, y_train, y_test = train_test_split(
X_train, Y_train, test_size=test_size, random_state=42)
data_train = lgb.Dataset(x_train, y_train, feature_name=feature_name)
data_test = lgb.Dataset(x_test, y_test, feature_name=feature_name)
evals_result = {}
gbm = lgb.train(model_param,
data_train,
num_boost_round=10000,
evals_result=evals_result,
valid_sets=[data_test],
early_stopping_rounds=50)
y_train_pred = gbm.predict(x_train)
train_acc = roc_auc_score(y_train, y_train_pred)
y_test_pred = gbm.predict(x_test)
test_acc = roc_auc_score(y_test, y_test_pred)
logger.info("lightgbm model train auc:" + str(train_acc) +
" test auc:" + str(test_acc))
return gbm, evals_result
except:
logger.error("light/gbm model train fail.")
return False, False
def get_model_train_result(evals_result,
model_name="default",
outputpath="./"):
'''
画出训练结果函数
:param evals_result:
:param model_name:
:param outputpath:
:return:
'''
try:
outputpath = outputpath + model_name + "_train_result.png"
ax = lgb.plot_metric(evals_result,
metric='binary_logloss',
figsize=(20, 13))
plt.savefig(outputpath)
except:
logger.error("create model train result fail.")
#raise RuntimeError("create model train result fail.")
return False
else:
logger.info("create model train result sucess.")
return True