【CV作业06-3】训练机器学习模型，预测肺结节良恶性

yierrrrr

已于 2023-06-30 20:24:28 修改

阅读量358

点赞数 1

分类专栏： CV作业文章标签：机器学习 python 计算机视觉图像处理

于 2023-04-13 22:38:04 首次发布

本文链接：https://blog.csdn.net/m0_71268863/article/details/130142025

版权

CV作业专栏收录该内容

8 篇文章 0 订阅

订阅专栏

读取数据

df = pd.read_csv('features.csv')
samples = np.array(df)

Z-score标准化处理
不对第一列和最后一列处理

samples[:, 1:-1] = preprocessing.scale(samples[:, 1:-1])

划分特征和标签

xs = samples[:, 1:-1]  # 特征
ys = samples[:, -1]  # 标签
ys = ys.astype(int)  # 不加这一步会在for fold, (train, test) in enumerate(five_folds):这一步报错

定义空列表来存储每个fold的性能指标和ROC曲线数据

accuracy_list = []
precision_list = []
recall_list = []
specificity_list = []
fpr_list = []
tpr_list = []

5折交叉划分数据集

skfold = StratifiedKFold(n_splits=5, shuffle=True)
five_folds = skfold.split(xs, ys)

模型训练

for fold, (train, test) in enumerate(five_folds):
    # 模型训练
    print('==========fold:', fold+1, '==========')
    mlp = MLP(hidden_layer_sizes=(64, 128, 256, 512),
              solver='sgd',
              batch_size=10,
              max_iter=200,
              learning_rate_init=0.001,
              verbose=True)
    mlp.fit(xs[train], ys[train])
    # 绘制损失函数曲线
    plt.plot(mlp.loss_curve_)
    img_path = './img/fold{}_loss.png'.format(fold+1)
    plt.xlabel('Iteration')
    plt.ylabel('loss')
    plt.title('Loss Curve')
    plt.savefig(img_path)

    # 模型评价
    y_pred = mlp.predict(xs[test])
    tn, fp, fn, tp = confusion_matrix(ys[test], y_pred).ravel()
    accuracy = (tp + tn) / (tp + tn + fp + fn)
    precision = tp / (tp + fp)
    recall = tp / (tp + fn)
    specificity = tn / (tn + fp)
    accuracy_list.append(accuracy)
    precision_list.append(precision)
    recall_list.append(recall)
    specificity_list.append(specificity)

    # 计算ROC曲线的FPR和TPR并将其添加到ROC曲线数据列表中
    fpr, tpr, _ = roc_curve(ys[test], y_pred)
    fpr_list.append(fpr)
    tpr_list.append(tpr)

计算各指标的平均值

mean_accuracy = np.mean(accuracy_list)
mean_precision = np.mean(precision_list)
mean_recall = np.mean(recall_list)
mean_specificity = np.mean(specificity_list)
mean_fpr = fpr
mean_tpr = tpr
mean_fpr[1] = np.mean([val[1] for val in fpr_list])
mean_tpr[1] = np.mean([val[1] for val in tpr_list])
auc = auc(mean_fpr, mean_tpr)

绘制 ROC曲线

plt.plot(mean_fpr, mean_tpr, 'k--', lw=2)
plt.xlim([-0.05, 1.05])  # 设置x、y轴的上下限，以免和边缘重合，更好的观察图像的整体
plt.ylim([-0.05, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('ROC Curve')