XGBoost feature_importances_ 参数返回 nan

Question

我有以下代码

xgb = XGBRegressor(booster='gblinear', reg_lambda=0, learning_rate=0.028) 

print(xgb)

xgb.fit(X_train_sc, y_train)

y_pred = xgb.predict(X_test_sc)

print("\nFeature Importances:")

for item in zip(feature_list_transform, xgb.feature_importances_):

    print("{1:10.4f} - {0}".format(item[0],item[1]))

print("\nR-squared, training set:")

print(xgb.score(X_train_sc,y_train))

print("R-squared, test set:")

print(xgb.score(X_test_sc,y_test))

print("\nRoot-mean squared error, from metrics:")

mse = mean_squared_error(y_test, y_pred)

rmse = np.sqrt(mse)

print(rmse)

输出是：

    XGBRegressor(base_score=0.5, booster='gblinear', colsample_bylevel=1,

           colsample_bytree=1, gamma=0, learning_rate=0.028, max_delta_step=0,

           max_depth=3, min_child_weight=1, missing=None, n_estimators=100,

           n_jobs=1, nthread=None, objective='reg:linear', random_state=0,

           reg_alpha=0, reg_lambda=0, scale_pos_weight=1, seed=None,

           silent=True, subsample=1)

    Feature Importances:

           nan - fertility_rate_log

           nan - life_expectancy_log

           nan - avg_supply_of_protein_of_animal_origin_log

           nan - access_to_improved_sanitation_log

           nan - access_to_improved_water_sources_log

           nan - obesity_prevalence_log

           nan - open_defecation_log

           nan - access_to_electricity_log

           nan - cereal_yield_log

           nan - population_growth_log

           nan - avg_value_of_food_production_log

           nan - gross_domestic_product_per_capita_ppp_log

           nan - net_oda_received_percent_gni_log

           nan - adult_literacy_rate

           nan - school_enrollment_rate_female

           nan - school_enrollment_rate_total

           nan - caloric_energy_from_cereals_roots_tubers

           nan - anemia_prevalence

           nan - political_stability

和错误：

c:\python36\lib\site-packages\xgboost\sklearn.py:420: RuntimeWarning: 在 true_divide 中遇到无效值 return all_features / all_features.sum()

如何修复这个nan并获得系数？最后，该模型运行良好。