Python 构建 Random Forest 和 XGBoost_randomforestclassifier(random_state=1,criterion= "

核心函数

def train_model(x_train, y_train, x_test, model_name, cv_state=True):
    '''
    Parameters
    ----------
    x_train : 训练集 x, np.array类型二维数组, [samples_train,features_train]
    y_train : 训练集 y  np.array类型一维数组, [samples_train]
    x_test  : 测试集 x  np.array类型二维数组, [samples_test,features_test]
    model_name : 选用什么模型 ：random forest / XGB
    cv_state: 是否需要模型因子挑选
    Returns
    -------
    y_predictoftest: 根据测试集 x 得到的预报测试集y
    '''
    if cv_state: #cv_state=True 则使用gridsearchCv 挑选最优参数
        if model_name == 'random forest':
            model = RandomForestRegressor(random_state=1, criterion='squared_error')
            # 挑选参数
            paremeters = [{"max_features": range(1, 32, 3),
                           "min_samples_leaf": range(1, 20, 3),
                           "max_depth": range(1, 20, 3)
                           }]
            grid = GridSearchCV(model, paremeters, cv=10, scoring="neg_mean_squared_error",verbose=10)
            grid.fit(x_train, y_train)
            print('best_params_=', grid.best_params_)
            print('best_score_=', grid.best_score_)
            model = RandomForestRegressor(random_state=1, criterion='mse',
                                          max_features=grid.best_params_['max_features'],
                                          min_samples_leaf=grid.best_params_['min_samples_leaf'],
                                          max_depth=grid.best_params_['max_depth'])
        elif model_name == 'XGB':
            model = xgb.XGBRegressor(random_state=1)
            # 挑选参数
            parameters = [{"eta": [0.3, 0.2, 0.1],
                           "max_depth": [3, 5, 6, 10, 20],
                           "n_estimators": [100, 200, 500],
                           'gamma': [0, 0.1, 0.2, 0.5, 1]
                           }]
            grid = GridSearchCV(model, parameters, cv=10, scoring="neg_mean_squared_error",verbose=10)
            grid.fit(x_train, y_train)
            print('best_params_=', grid.best_params_)
            print('best_score_=', grid.best_score_)
            model = xgb.XGBRegressor(random_state=1,
                                     eta=grid.best_params_['eta'],
                                     max_depth=grid.best_params_['max_depth'],
                                     n_estimators=grid.best_params_['n_estimators'],
                                     gamma=grid.best_params_['gamma'])
 
    else: #cv_state=False 则根据自己需要修改模型参数后直接推理
        if model_name == 'random forest':
            model = RandomForestRegressor(random_state=1, criterion='mse', max_depth=7, max_features=31,
                                          min_samples_leaf=10)  # random_state=1,criterion='mse',max_depth=7,max_features=31,min_samples_leaf=10
        elif model_name == 'XGB':
            #model = xgb.XGBRegressor(random_state=1, learning_rate=0.1, max_depth=2, n_estimators=100)
            model = xgb.XGBRegressor(random_state=1, gamma=0.1, max_depth=3, n_estimators=100)
    regr = model.fit(x_train, y_train)
    y_predictoftest = regr.predict(x_test)
 
    return y_predictoftest

在调用train_model前你需要额外做的一些事情：

1. 加载必要的库：

import numpy as np
import pandas as pd
from sklearn.model_selection import GridSearchCV
from sklearn.ensemble import RandomForestRegressor
import xgboost as xgb
from sklearn.model_selection import StratifiedKFold

2. 根据自己的数据处理数据

x_train,y_train, x_test，y_test, 类型描述可看train_model中描述

3. 调用函数

y_predictoftest=train_model(x_train,y_train, x_test,'XGB',cv_state=True)

如果想调用XGB，就用'XGB'；

如果想调用random forest，就用'random forest'

cv_state: 是否需要GridSearchCV进行调参。

4. 如果想自己调参，可以参阅官网

scikit-learn 官网Random Forest 部分：

分类器：

sklearn.ensemble.RandomForestClassifier — scikit-learn 1.1.1 documentation

回归器：

sklearn.ensemble.RandomForestRegressor — scikit-learn 1.1.1 documentation

XGBoost ：

XGBoost Parameters — xgboost 1.6.1 documentation