GridSearchCV——信用卡违约率分析示例

一、简介

GridSearchCV，即网格搜索，给定指定的模型、参数，GridSearchCV可以自动的寻找到最优的模型及参数。

二、代码

引用包：

from sklearn.model_selection import GridSearchCV

构造分类器：

classifiers = [
    SVC(random_state = 1, kernel = 'rbf'),    
    DecisionTreeClassifier(random_state = 1, criterion = 'gini'),
    RandomForestClassifier(random_state = 1, criterion = 'gini'),
    KNeighborsClassifier(metric = 'minkowski'),
]

分类器参数：

classifier_param_grid = [
            {'svc__C':[1], 'svc__gamma':[0.01]},
            {'decisiontreeclassifier__max_depth':[6, 9, 11]},
            {'randomforestclassifier__n_estimators':[3, 5, 6]} ,
            {'kneighborsclassifier__n_neighbors':[4, 6, 8]},
]

完整代码：

# -*- coding: utf-8 -*-
# 信用卡违约率分析
import pandas as pd
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import accuracy_score
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from matplotlib import pyplot as plt
import seaborn as sns

# 数据加载
data = pd.read_csv('./UCI_Credit_Card.csv')
# 数据探索
print(data.shape) # 查看数据集大小
print(data.describe()) # 数据集概览

# 查看下一个月违约率的情况
next_month = data['default.payment.next.month'].value_counts()
print(next_month)
df = pd.DataFrame({'default.payment.next.month': next_month.index,'values': next_month.values})

plt.rcParams['font.sans-serif'] = ['SimHei'] #用来正常显示中文标签
plt.figure(figsize = (6,6))
plt.title('信用卡违约率客户\n (违约：1，守约：0)')
sns.set_color_codes("pastel")
sns.barplot(x = 'default.payment.next.month', y="values", data=df)
plt.show()

# 特征选择，去掉ID字段、最后一个结果字段即可
data.drop(['ID'], inplace=True, axis =1) #ID这个字段没有用
target = data['default.payment.next.month'].values

columns = data.columns.tolist()
columns.remove('default.payment.next.month')
features = data[columns].values

# 30%作为测试集，其余作为训练集
train_x, test_x, train_y, test_y = train_test_split(features, target, test_size=0.30, stratify = target, random_state = 1)
    
# 构造各种分类器
classifiers = [
    SVC(random_state = 1, kernel = 'rbf'),    
    DecisionTreeClassifier(random_state = 1, criterion = 'gini'),
    RandomForestClassifier(random_state = 1, criterion = 'gini'),
    KNeighborsClassifier(metric = 'minkowski'),
]

# 分类器名称
classifier_names = [
            'svc', 
            'decisiontreeclassifier',
            'randomforestclassifier',
            'kneighborsclassifier',
]

# 分类器参数
classifier_param_grid = [
            {'svc__C':[1], 'svc__gamma':[0.01]},
            {'decisiontreeclassifier__max_depth':[6, 9, 11]},
            {'randomforestclassifier__n_estimators':[3, 5, 6]} ,
            {'kneighborsclassifier__n_neighbors':[4, 6, 8]},
]
 
# 对具体的分类器进行GridSearchCV参数调优
def GridSearchCV_work(pipeline, train_x, train_y, test_x, test_y, param_grid, score = 'accuracy'):
    response = {}
    gridsearch = GridSearchCV(estimator = pipeline, param_grid = param_grid, scoring = score)
    # 寻找最优的参数 和最优的准确率分数
    search = gridsearch.fit(train_x, train_y)
    print("GridSearch最优参数：", search.best_params_)
    print("GridSearch最优分数： %0.4lf" %search.best_score_)
    predict_y = gridsearch.predict(test_x)
    print("准确率 %0.4lf" %accuracy_score(test_y, predict_y))
    response['predict_y'] = predict_y
    response['accuracy_score'] = accuracy_score(test_y,predict_y)
    return response
 
    
for model, model_name, model_param_grid in zip(classifiers, classifier_names, classifier_param_grid):
    pipeline = Pipeline([
            ('scaler', StandardScaler()),
            (model_name, model)
    ])
    result = GridSearchCV_work(pipeline, train_x, train_y, test_x, test_y, model_param_grid , score = 'accuracy')
    

运行结果：

GridSearch最优参数： {'svc__C': 1, 'svc__gamma': 0.01}
GridSearch最优分数： 0.8186
准确率 0.8172
GridSearch最优参数： {'decisiontreeclassifier__max_depth': 6}
GridSearch最优分数： 0.8208
准确率 0.8113
GridSearch最优参数： {'randomforestclassifier__n_estimators': 6}
GridSearch最优分数： 0.8004
准确率 0.7994
GridSearch最优参数： {'kneighborsclassifier__n_neighbors': 8}
GridSearch最优分数： 0.8040
准确率 0.8036