x_train, x_test, y_train, y_test = train_test_split(feature_df, label_df, test_size=0.33, random_state=42)
train_data = lgb.Dataset(data=x_train,label=y_train)
test_data = lgb.Dataset(data=x_test,label=y_test)
param = {'max_depth':10, 'objective':'binary','num_threads':8,
'learning_rate':0.1,'bagging':0.7,'feature_fraction':0.7,
'lambda_l1':0.1,'lambda_l2':0.2,'seed':123454}
param['metric'] = ['auc']
bst = lgb.train(param, train_data,num_boost_round=150,early_stopping_rounds=100, valid_sets=[test_data])
bst.save_model("lgb.model")
y_train_binary = bst.predict(x_train, num_iteration=bst.best_iteration) # type:np.numarray
y_pred_binary = bst.predict(x_test, num_iteration=bst.best_iteration) # type:np.numarray
重要度