[DA45] 使用SVM进行乳腺癌检测

 # 读取并探索数据

    
 inputfile = './data.csv'
    
 data = pd.read_csv(inputfile)
    
 print(data.head())
    
 print(data.describe())
    
 print(data.info())
    
 print(data.diagnosis.value_counts())
    
  
    
 # 数据清洗与分类
    
 # print(data.columns)
    
 data_y = data.diagnosis.replace(['B', 'M'], [0, 1])
    
 # print(data_y.value_counts())
    
 # 把平均值数据, 方差数据与最坏值数据分开
    
 data_mean = data[data.columns[2:12]]
    
 # print(data_mean.columns)
    
 data_se = data[data.columns[12:22]]
    
 # print(data_se.columns)
    
 data_worst = data[data.columns[22:32]]
    
 # print(data_worst.columns)

 # 1. 相关性分析降维

    
 # 画相关性系数图
    
 mean_corr = data_mean.corr()
    
 sns.heatmap(mean_corr, annot=True)
    
 plt.show()
    
 # 特征选择
    
 features1 = data[['radius_mean', 'texture_mean', 'smoothness_mean',
    
               'compactness_mean', 'symmetry_mean', 'fractal_dimension_mean']]
    
 # 标准化
    
 ss = StandardScaler()
    
 features1 = ss.fit_transform(features1)
    
 # 训练集数据集分割
    
 train_x, test_x, train_y, test_y = train_test_split(features1, data_y, test_size=.3, random_state=33)
    
 # LinearSVC模型
    
 model1 = LinearSVC()
    
 model1.fit(train_x, train_y)
    
 pre_1 = model1.predict(test_x)
    
 print('模型1得分:', accuracy_score(test_y, pre_1))

 # 2. 主成分分析降维

    
 # PCA 模型
    
 pca = PCA(n_components=6)
    
 # pca = PCA(n_components='mle')
    
 model2 = Pipeline([
    
     ('StandardScaler', StandardScaler()),
    
     ('PCA', pca)])
    
 features2 = model2.fit_transform(data_mean)
    
 print(features2.shape)
    
 print('贡献率:', pca.explained_variance_ratio_)
    
 # 训练数据集分割
    
 train_x, test_x, train_y, test_y = train_test_split(features2, data_y, test_size=.3, random_state=33)
    
 # LinearSVC模型
    
 model2 = LinearSVC()
    
 model2.fit(train_x, train_y)
    
 pre_2 = model2.predict(test_x)
    
 print('模型2得分:', accuracy_score(test_y, pre_2))