【Sklearn】【实战】用SVM进行乳腺癌检测，模拟线上部署（一）

0. 前言

• 转载请注明出处
• 关于本文所用的SVC API 的详细解释可以参考这篇文章：【Sklearn】【API详解】【SVM】sklearn.svm.SVC参数详解（一）
• github代码地址为：shinylover/Datamining-1_Stars-SVM_Breast_Cancer

1. 需求

1. 通过sklearn中的自带乳腺癌数据集 训练 svc分类器
2. 对模型进行保存，即为模型持久化
3. 载入训练好的模型，对模型模拟线上测试

2. 数据集介绍

1. 数据集官网介绍：sklearn.datasets.load_breast_cancer
2. 二分类任务：212（M） + 357（B） = 569
3. 特征个数：30 维
   

3. 导入依赖包

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.svm import SVC
from sklearn.metrics import  classification_report
from sklearn.model_selection import GridSearchCV
import joblib
import time

4. 载入数据集

cancer = load_breast_cancer()
X = cancer.data
Y = cancer.target
print(f'The shape of cancer dataset is {X.shape}')
print(f'The shape of label is {Y.shape}')

5. 数据预处理

• 导入数据和标签

cancer = load_breast_cancer()
X = cancer.data
Y = cancer.target
print(f'The shape of cancer dataset is {X.shape}')
print(f'The shape of label is {Y.shape}')

• 运行结果为
  

6. 特征工程

• 由于此数据集的特征是构造好的，我们无需构造，所以此步骤省略

7. 建模

1. 划分训练集合测试集，比例为 8:2

# 划分训练集合测试集，比例为 8:2
X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.2)
print(f'The shape of X_train is {X_train.shape}')
print(f'The shape of y_train is {y_train.shape}')
print(f'The shape of X_test is {X_test.shape}')
print(f'The shape of y_test is {y_test.shape}')

2. 运行结果为
   
3. 建模：SVC介绍为：sklearn.svm.SVC

# 建模
clf_svc = SVC()
clf_svc.fit(X_train, y_train)

8. 模型评估

1. 我们使用两种评估方式，一种是只有准确率，另外一种是分类报告，包括精确度，召回率，F1分等
2. API介绍：sklearn.metrics.classification_report
3. 代码如下

# 模型评估
accuracy = clf_svc.score(X_test, y_test)
print(f'The accuracy of the SVC model is {accuracy*100}%')
y_pred = clf_svc.predict(X_test)
print(f'The report of the SVC model is \n {classification_report(y_test, y_pred)}')

4. 运行结果
   

9. 参数调优

1. 我们使用了网格搜索法进行参数调优
2. API介绍网站为：sklearn.model_selection.GridSearchCV
3. 代码如下

# 参数调优
parameters = {'kernel':('linear', 'rbf'), 'C':[1, 10]}
svc = SVC()
clf = GridSearchCV(svc, parameters)
clf.fit(X_train, y_train)
print(f'The best parameters are {clf.best_params_}')
accuracy = clf.score(X_test, y_test)
print(f'The accuracy of the SVC model is {accuracy*100}%')
y_pred = clf.predict(X_test)
print(f'The report of the SVC model is \n {classification_report(y_test, y_pred)}')
# 保存模型
joblib.dump(clf, "svc_breast_cancer.m")

4. 运行结果
   
5. 由此可见，我们最佳参数为，C=10， kernel =“linear”， 准确率也从 86% 提升到了 98%

10. 模拟线上部署

1. 这段我们通过time.sleep() 来模拟模型的真实线上部署情况
2. 代码如下

# 模拟线上部署
clf_online = joblib.load("svc_breast_cancer.m")
for i, e in enumerate(X_test):
    result_SVM = clf_online.predict([e])
    print(f"--------------The {i+1}th data result is: ---------------------- ")
    print(f"The Result of Model is {result_SVM[0]}")
    print(f"The True value is {y_test[i]} \n")
    time.sleep(1)

3. 运行结果
   

11. 参考资料

1. 数据集官网介绍：sklearn.datasets.load_breast_cancer
2. SVC介绍为：sklearn.svm.SVC
3. API介绍：sklearn.metrics.classification_report
4. API介绍网站为：sklearn.model_selection.GridSearchCV