准确度、精确度、召回率、ROC曲线、AUC值

在Python中，特别是在机器学习库如scikit-learn中，评估模型性能通常涉及以下几个步骤： 1. **计算AUC (Area Under the Curve)**: ```python from sklearn.metrics import roc_auc_score y_true = ... # 真实标签 y_pred_proba = ... # 预测概率 auc = roc_auc_score(y_true, y_pred_proba) ``` 2. **准确度 (Accuracy)**: ```python accuracy = accuracy_score(y_true, y_pred) # y_pred 是预测的类别 ``` 3. **精确度 (Precision)**: ```python precision = precision_score(y_true, y_pred) ``` 4. **召回率 (Recall, Sensitivity or True Positive Rate)**: ```python recall = recall_score(y_true, y_pred) ``` 5. **F1-score**: ```python f1 = f1_score(y_true, y_pred) ``` 6. **绘制ROC曲线 (Receiver Operating Characteristic curve)**: ```python import matplotlib.pyplot as plt from sklearn.metrics import plot_roc_curve plot_roc_curve(model, X_test, y_test) plt.show() ``` 7. **混淆矩阵 (Confusion Matrix)**: ```python conf_mat = confusion_matrix(y_true, y_pred) print(conf_mat) ``` 以上代码假设`y_true`是实际标签列表，`y_pred`是预测结果，而`model`是你训练好的分类模型，`X_test`和`y_test`是测试集数据。