def gb_train(self,X_train,y_train,X_val,y_val):
        self.logger.info("..creating the datasets for gb")
        param_grid = {
            'num_leaves': [20],
            'min_data_in_leaf': [10],
            'max_depth': [40],
            'bagging_fraction': [0.85],
            'colsample_bytree': [0.75],
            'subsample': [0.7],
            'reg_alpha': [1.2]}
        std_params = {'boosting_type':'gbdt',
                      'objective':'regression',
                      'bagging_freq': 5,
                      'learning_rate':0.05,
                      'metric':'rmse',
                      'num_leaves': 10,
                      'min_data_in_leaf': 1,
                      'max_depth': 15,
                      'bagging_fraction': 0.95}
        lgb_estimator = lgb.LGBMRegressor(boosting_type='gbdt',
                                          objective='regression',
                                          bagging_freq=5,
                                          num_boost_round= 10000,
                                          learning_rate= 0.075,
                                          metric='rmse',
                                          early_stopping_rounds=5,
                                          verbose=-2)
        gsearch = GridSearchCV(estimator=lgb_estimator,
                               param_grid=param_grid,
                               cv=4,
                               refit=True)
        lgb_model = gsearch.fit(X = X_train,
                                y = y_train,
                                eval_set = (X_val, y_val),
                                eval_metric = 'rmse',
                                verbose=-3)
        self.logger.info("..cross validation has finished correctly")
        msg = "..the best score over cross validation is {}"
        msg = msg.format(lgb_model.best_score_)
        self.logger.info(msg)
        self.logger.info("..showing best params")
        params = lgb_model.best_params_
        self.logger.info("..{}".format(params))
        for key in std_params:
            if key in params:
                std_params[key]=params[key]
        lgb_X_train = lgb.Dataset(X_train,y_train)
        lgb_X_eval = lgb.Dataset(X_val,y_val)
        gbm = lgb.train(std_params,
                        num_boost_round=10000,
                        early_stopping_rounds=5,
                        train_set=lgb_X_train,
                        valid_sets=lgb_X_eval)
        self.logger.info("..plotting features importance")
        lgb.plot_importance(gbm)
        #plt.show()
        gbm.save_model('gbm_model.txt')
        self.gb_model = gbm