机器学习——常用性能指标及sklearn中的模型评估

机器学习专栏：机器学习专栏

导入方式:sklearn.metrics

from sklearn.metrics import mean_squared_error
from sklearn.metrics import r2_score

一、回归指标

1. 回归方差(反应自变量与因变量之间的相关程度)

explained_variance_score(y_true, y_pred, sample_weight=None, multioutput=‘uniform_average’)

2. 平均绝对误差

mean_absolute_error(y_true,y_pred,sample_weight=None,multioutput=‘uniform_average’)

3. 均方差

mean_squared_error(y_true,y_pred,sample_weight=None,multioutput=‘uniform_average’)

4. 中值绝对误差

median_absolute_error(y_true, y_pred) 

5. R平方值

r2_score(y_true,y_pred,sample_weight=None,multioutput=‘uniform_average’) 

二、分类指标

1. 精度

accuracy_score(y_true,y_pre)

2. ROC曲线下的面积：较大的AUC代表了较好的performance

auc(x, y, reorder=False)

3. 根据预测得分计算平均精度(AP)

average_precision_score(y_true, y_score, average=‘macro’, sample_weight=None)

4. 通过计算混淆矩阵来评估分类的准确性，返回混淆矩阵

confusion_matrix(y_true, y_pred, labels=None, sample_weight=None)

5. F1值

F1 = 2 * (precision * recall) / (precision + recall)
precision(查准率)=TP/(TP+FP)
recall(查全率)=TP/(TP+FN)

precision_score(y_true, y_pred, labels=None, pos_label=1, average=‘binary’)

recall_score(y_true, y_pred, labels=None, pos_label=1, average=‘binary’, sample_weight=None)

f1_score(y_true, y_pred, labels=None, pos_label=1, average=‘binary’, sample_weight=None)

6. 对数损耗，又称逻辑损耗或交叉熵损耗

log_loss(y_true, y_pred, eps=1e-15, normalize=True, sample_weight=None, labels=None)

7. 计算ROC曲线下的面积就是AUC的值，the larger the better

roc_auc_score(y_true, y_score, average=‘macro’, sample_weight=None)

8. ROC曲线

ROC曲线越接近左上角，代表模型越好，即ACU接近1

def plot_roc(labels, predict_prob):
    false_positive_rate, true_positive_rate, thresholds = roc_curve(labels, predict_prob)
    roc_auc = auc(false_positive_rate, true_positive_rate)
    plt.title('ROC')
    plt.plot(false_positive_rate, true_positive_rate,'b',label='AUC = %0.4f'% roc_auc)
    plt.legend(loc='lower right')
    plt.plot([0,1],[0,1],'r--')
    plt.ylabel('TPR')
    plt.xlabel('FPR')
    plt.show()