gridSearch做参数调优的练习 将K近邻 svm 决策树 randomforest adaboost模型的优化全都放进来

代码实现参考了这篇文章 https://blog.csdn.net/weixin_41171061/article/details/83859856

比较各种组合下分类效果最好的一个方案

用鸢尾花数据

#!/usr/bin/python
# -*- coding:utf-8 -*-
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import accuracy_score
from sklearn.metrics import auc
import numpy as np
import matplotlib as mpl
import pandas as pd
seed = 1231
np.random.seed(seed)
names = ['KNearesNeighbors','SVC','Decision Tree', 'Random Forest','AdaBoostClassifier']
classifiers = [KNeighborsClassifier(),SVC(kernel='rbf'),DecisionTreeClassifier(), RandomForestClassifier(),AdaBoostClassifier(base_estimator=DecisionTreeClassifier())]
parameter_knn = {'n_neighbors':[3,5]}
parameter_svc={'C': np.logspace(-2, 2, 10), 'gamma': np.logspace(-2, 2, 10)}
parameter_dtc = {'max_features': ['auto', 'sqrt', 'log2', None], 'max_depth': range(3, 100, 2)}
parameter_rfc = {'n_estimators': range(5, 200, 40), 'max_features': ['auto', 'sqrt', 'log2', None],
                 'max_depth': range(3, 100, 20)}
parameter_adb = {"base_estimator__criterion": ["gini", "entropy"],
          "base_estimator__splitter":   ["best", "random"],
          "n_estimators": [3, 5, 7]}

parameters = [parameter_knn,parameter_svc,parameter_dtc, parameter_rfc, parameter_adb]

def gird_search_model(clf, param, name, x_train, y_train, x_test,y_test):  # clf-classifier;param-parameter;name-classifier_name
    model = GridSearchCV(clf, param, cv=5)  # GridSearchCV模型
    fit = model.fit(x_train, y_train)  # GridSearchCV模型拟合训练集数据，并返回训练器集合为fit
    y_train_pred = fit.best_estimator_.predict(x_train)  # 用训练器集合中最好的estimator预测y_train_pred
    y_test_pred = fit.best_estimator_.predict(x_test)  # 用训练器集合中最好的estimator预测y_test_pred
    print('MODEL : %r' % name)
    print('Best cv_test_roc_auc: %f using %s' % (
    fit.best_score_, fit.best_params_))  # 训练器集合fit中最好的模型得到的：best_score和best_params

    train_score_list = []
    test_score_list = []
    score_list = []
    model_metrics_name = [accuracy_score]  # 模型评价指标，与scoreing相对应
    for matrix in model_metrics_name:  # 计算各个模型评价指标
        train_score = matrix(y_train, y_train_pred)  # 计算训练集的
        test_score = matrix(y_test, y_test_pred)  # 计算测试集的
        train_score_list.append(train_score)  # 把训练集的各个模型指标放在同一行
        test_score_list.append(test_score)  # 把测试集的各个模型指标放在同一行
    score_list.append(train_score_list)  # 合并训练集和测试集的结果（便于展示）
    score_list.append(test_score_list)  # 合并训练集和测试集的结果（便于展示）
    score_df = pd.DataFrame(score_list, index=['train', 'test'],
                            columns=['accuracy'])  # 将结果显示为df格式，加上行列index
    print('EVALUATE_METRICS:')
    print(score_df)
    return score_list, y_train_pred, y_test_pred


if __name__ == "__main__":
    mpl.rcParams['font.sans-serif'] = [u'SimHei']
    mpl.rcParams['axes.unicode_minus'] = False

    iris_feature = '花萼长度', '花萼宽度', '花瓣长度', '花瓣宽度'
    path = 'iris.data'  # 数据文件路径
    data = pd.read_csv(path, header=None)
    x_prime = data.iloc[:,0:3]
    y = pd.Categorical(data[4]).codes
    train_score_list = []
    test_score_list = []
    x_train,x_test,y_train,y_test=train_test_split(x_prime, y, train_size=0.7, random_state=0)
    for clf,param,name in zip(classifiers,parameters,names):
        score_list,y_train_pred,y_test_pred =gird_search_model(clf,param,name,x_train,y_train,x_test,y_test)
        train_score_list.append(score_list[0])
        test_score_list.append(score_list[1])
        print('------------------------------------------------------------------------------------')
    train_score_df = pd.DataFrame(train_score_list,index=names,columns=['acc'])
    test_score_df = pd.DataFrame(test_score_list,index=names,columns=['acc'])
    print('TRAIN_SCORE:')
    print(train_score_df)
    print('TEST_SCORE:')
    print(test_score_df)

输出最优的参数及模型

MODEL : 'KNearesNeighbors'
Best cv_test_roc_auc: 0.961905 using {'n_neighbors': 5}
EVALUATE_METRICS:
       accuracy
train  0.961905
test   0.977778
--------------------------------------------------------------------------------------------------------
MODEL : 'SVC'
Best cv_test_roc_auc: 0.980952 using {'C': 1.6681005372000592, 'gamma': 0.21544346900318834}
EVALUATE_METRICS:
       accuracy
train  0.980952
test   0.955556
--------------------------------------------------------------------------------------------------------
MODEL : 'Decision Tree'
Best cv_test_roc_auc: 0.952381 using {'max_depth': 3, 'max_features': None}
EVALUATE_METRICS:
       accuracy
train  0.961905
test   0.911111
--------------------------------------------------------------------------------------------------------
MODEL : 'Random Forest'
Best cv_test_roc_auc: 0.952381 using {'max_depth': 3, 'max_features': None, 'n_estimators': 85}
EVALUATE_METRICS:
       accuracy
train  0.961905
test   0.911111
--------------------------------------------------------------------------------------------------------
MODEL : 'AdaBoostClassifier'
Best cv_test_roc_auc: 0.933333 using {'base_estimator__criterion': 'gini', 'base_estimator__splitter': 'best', 'n_estimators': 5}
EVALUATE_METRICS:
       accuracy
train  1.000000
test   0.955556
--------------------------------------------------------------------------------------------------------
TRAIN_SCORE:
                         acc
KNearesNeighbors    0.961905
SVC                 0.980952
Decision Tree       0.961905
Random Forest       0.961905
AdaBoostClassifier  1.000000

TEST_SCORE:
                         acc
KNearesNeighbors    0.977778
SVC                 0.955556
Decision Tree       0.911111
Random Forest       0.911111
AdaBoostClassifier  0.955556