【机器学习】基于多变量特征的乳腺癌自动诊断

    # 读取数据
    import numpy as np
    import pandas as pd
    from sklearn import preprocessing
    # 原始数据以csv格式读取
    dataset = pd.read_csv(r'C:\Users\Lenovo\Desktop\analyse\wdbc.data.csv',header=None)
    # 标签化diagnosis列
    encoder = preprocessing.LabelEncoder().fit(dataset['diagnosis'])
    dataset['diagnosis'] = encoder.transform(dataset['diagnosis'])
    print('标签: %s' % encoder.classes_)

    # 获取数据
    X = np.array(dataset)
    dataset_target = X[:, 1] # 原始标签数据
    dataset_data = X[:, 2:32] # 原始输入数据
    # 数据集计数
    from collections import Counter
    Counter(dataset_target)
    # Counter({1.0: 212, 0.0: 357})

    # SMOTE过采样，平衡正负样本数据量，提升识别效果
    from imblearn.over_sampling import SMOTE
    smo = SMOTE()
    # 得到采样后的数据样本
    dataset_data_sm, dataset_target_sm = smo.fit_resample(dataset_data, dataset_target)
    Counter(dataset_target_sm)

    from sklearn import preprocessing
    import numpy as np
    dataset_data_scaled = preprocessing.scale(dataset_data_sm)

    # 通过网格搜索法得出最优的参数组合
    # 随机森林是由多个决策树集成而成。n_estimators为弱学习器的个数，n_estimators太小，容易欠拟合，太大，容易过拟合。max_depth决定了决策树的最大深度。
    from sklearn.model_selection import GridSearchCV
    
    param =  {'n_estimators': np.arange(2, 20, 2),'max_depth': np.arange(2, 11)}
    gc = GridSearchCV(model, param_grid=param, cv=5)
    gc.fit(X_train,y_train)
    print("最佳参数组合：",gc.best_params_)
    depth = gc.best_params_['max_depth']
    estimator = gc.best_params_['n_estimators']
    # 最佳参数组合： {'max_depth': 7, 'n_estimators': 16}
    # 确定n_estimators和max_depth参数，调整其他两个参数
    # max_features的增加一般能提升模型性能，在每个节点，选择更多。但是会降低算法的速度，需要适当平衡选择最佳的值。
    # criterion为CART树做划分时对特征的评价标准，分类模型和回归模型的损失函数是不一样的
    model = RandomForestClassifier(random_state=90, max_depth=depth, n_estimators=estimator)
    param =  {"max_features":np.arange(2,30,2),"criterion":['gini','entropy']}
    gc = GridSearchCV(model, param_grid=param, cv=5)
    gc.fit(X_train,y_train)
    print("最佳参数组合：",gc.best_params_)
    criterion = gc.best_params_['criterion']
    feature = gc.best_params_['max_features']
    # 最佳参数组合： {'criterion': 'gini', 'max_features': 14}

    model = RandomForestClassifier(random_state=90, max_depth=depth, n_estimators=estimator,criterion=criterion,max_features=feature)
    model.fit(X_train, y_train)
    # 基于测试数据进行预测
    y_pred = model.predict(X_test)
    # 得到预测结果,四舍五入去掉小数点位
    predictions = [round(value) for value in y_pred] 
    # 评估准确率效果
    accuracy = accuracy_score(y_test, predictions)
    precision = precision_score(y_test, predictions)
    recall = recall_score(y_test, predictions)
    f1 = f1_score(y_test,predictions)
    tn, fp, fn, tp = confusion_matrix(y_test, y_pred,labels=[0,1]).ravel()
    fpr = fp/(fp+tn)
    spe = tn/(fp+tn)
    
    #准确率: 0.958041958041958
    #精度: 0.948051948051948
    #召回率: 0.9733333333333334
    #FPR： 0.058823529411764705
    #特异度： 0.9411764705882353
    #F1值: 0.9605263157894737