gridsearchcv参数_网格搜索寻找逻辑回归的最佳超参数

鸢尾花分类数据集iris.csv.

sklearn 0.21.3

iris.csv的部分数据如下图所示

import pandas as pd
import numpy as np
from sklearn import preprocessing
from sklearn.model_selection import train_test_split
from sklearn import linear_model
from sklearn import metrics
import matplotlib.pyplot as plt
from sklearn.model_selection import GridSearchCV
import seaborn as sns
import warnings

warnings.filterwarnings("ignore")

##设置控制台显示格式
np.set_printoptions(threshold=np.inf)       #显示所有行数据
pd.set_option('display.max_columns', None)  #显示所有列

#导入数据
datafile = "./iris.csv"
data_df = pd.read_csv(datafile, header=0)   #第一行是标题

#数据描述统计
print('data.shape','-'*30,'n',data_df.shape)
print('data.dtypes','-'*30,'n',data_df.dtypes)
print('data.describe','-'*30,'n',data_df.describe())
print('缺失值个数','-'*30,'n',data_df.isnull().sum())

#输出结果是 
#箱型图，查看特征分布情况
X_tmp = data_df.drop(columns=['label'])
X_tmp.plot.box()
plt.show()   

#从箱形图中可以看出， 没有异常值。
#柱状图，查看每个类别数量
sns.countplot(x="label", data=data_df)
plt.show() 
 
#特征预处理
feature_names = ["sepal_length","sepal_width","petal_length","petal_width"]
print("特征预处理之前 n",data_df)

min_max_scaler = preprocessing.MinMaxScaler()
data_df["sepal_length"] = min_max_scaler.fit_transform(data_df["sepal_length"].values.reshape(-1,1))
data_df["sepal_width"] = min_max_scaler.fit_transform(data_df["sepal_width"].values.reshape(-1,1))
data_df["petal_length"] = min_max_scaler.fit_transform(data_df["petal_length"].values.reshape(-1,1))
data_df["petal_width"] = min_max_scaler.fit_transform(data_df["petal_width"].values.reshape(-1,1))
print("特征预处理之后 n",data_df)


#拆分成属性X,标签y
X = data_df.drop(columns=['label'])
y = data_df['label']
#训练集和验证集
X_train, X_test, y_train, y_test = train_test_split(X,y)

#训练模型
logisticregression = linear_model.LogisticRegression(penalty='l2',C=1.0)
model =logisticregression.fit(X_train,y_train)
print('模型','-'*30,'n',model)
print('模型系数','-'*30,'n',model.coef_)
print('模型截距','-'*30,'n',model.intercept_)
print('模型得分','-'*30,'n',model.score(X_test,y_test))

#测试模型
y_predict = model.predict(X_test)#打印真实值与预测值
y_predict_df = pd.DataFrame(y_predict,columns=['y_predict'],index=y_test.index)
y_test_predict_df = pd.concat([y_test,y_predict_df],axis = 1)
print('真实值与预测值','-'*30,'n',y_test_predict_df)

#评估模型
#F1 = 2 * (precision * recall) / (precision + recall)
accuracy = metrics.accuracy_score(y_test, y_predict)
confusionmatrix = metrics.confusion_matrix(y_test, y_predict)
target_names = ['class 0', 'class 1', 'class 2']
classifyreport = metrics.classification_report(y_test, y_predict,target_names=target_names)
print('分类准确率 ',accuracy) # 混淆矩阵对角线元素之和/所有元素之和
print('混淆矩阵 n', confusionmatrix)
print('分类结果报告 n', classifyreport)

#模型优度的可视化展现
fpr, tpr, _ = metrics.roc_curve(y_test, y_predict,pos_label=2)
auc = metrics.auc(fpr, tpr)
plt.style.use('ggplot')   #设置绘图风格
plt.plot(fpr, tpr,'')    #绘制ROC曲线
plt.plot((0,1),(0,1),'r--')   #绘制参考线
plt.text(0.5,0.5,'AUC=%.2f' %auc)   #添加文本注释
plt.title('LogisticRegression ROC')  #设置标题
plt.xlabel('False Positive Rate')   #设置坐标轴标签
plt.ylabel('True Positive Rate')
plt.tick_params(top='off', right='off') #去除图形顶部边界和右边界的刻度
plt.show()   #图形显示

#优化模型,选择模型最佳参数
parameters = {'penalty': ('l1', 'l2'),'C': (0.01, 0.1, 1, 10)}
grid_search = GridSearchCV(linear_model.LogisticRegression(), parameters,  verbose=0, scoring='accuracy', cv=5)
grid =grid_search.fit(X_train, y_train)
print('最佳效果：%0.3f' % grid_search.best_score_)
best_parameters= grid_search.best_estimator_.get_params()
print('最佳参数：n', best_parameters)

最佳效果：0.946
最佳参数：
 {'C': 10, 'class_weight': None, 'dual': False, 'fit_intercept': True, 'intercept_scaling': 1, 'l1_ratio': None, 'max_iter': 100, 'multi_class': 'warn', 'n_jobs': None, 'penalty': 'l1', 'random_state': None, 'solver': 'warn', 'tol': 0.0001, 'verbose': 0, 'warm_start': False}