使用sklearn计算二进制数组和浮点分数的roc曲线时出现问题

Question

我正在尝试为这样的一组预测计算roc曲线

 fpr, tpr, thresholds = roc_curve(y_test, probas)

这是y_test数组

array([-10.54, -9.49, -9.4, -9.37, -9.36, -9.31, -9.28, -9.14, -9.11,
       -9.03, -9.01, -9.0, -8.99, -8.98, -8.96, -8.91, -8.9, -8.9, -8.9,
       -8.89, -8.88, -8.86, -8.86, -8.84, -8.83, -8.78, -8.76, -8.74,
       -8.74, -8.69, -8.69, -8.69, -8.67, -8.64, -8.61, -8.57, -8.51, -8.5,
       -8.49, -8.48, -8.4, -8.34, -8.33, -8.3, -8.29, -8.29, -8.27, -8.26,
       -8.25, -8.22, -8.15, -8.12, -8.1, -8.08, -8.04, -8.04, -7.96, -7.94,
       -7.94, -7.85, -7.83, -7.82, -7.82, -7.81, -7.76, -7.74, -7.71,
       -7.65, -7.57, -7.54, -7.47, -7.4, -7.39, -7.34, -7.33, -7.32, -7.27,
       -7.23, -7.16, -7.08, -7.05, -6.92, -6.9, -6.89, -6.86, -6.86, -6.83,
       -6.78, -6.73, -6.69, -6.59, -6.57, -6.4, -6.37, -6.21, -6.19, -6.16,
       -6.04, -6.04, -5.57, -5.54, -5.35, -5.24, -5.0, -4.92], dtype=object)

这是probas数组

array([1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], dtype=object)

现在，当我尝试跑步时

fpr, tpr, thresholds = roc_curve(y_test, probas)

我收到ValueError

-> 318提高ValueError（“不支持{0}格式” .format（y_type））319320 check_consistent_length（y_true，y_score，sample_weight）

ValueError：不支持连续格式

我该如何解决？

Answer 1

好像您切换了目标分数和二进制标签。 我必须从您的数组中删除dtype=object才能使其正常工作。 以下是可行的解决方案。 根据此处的官方页面， roc_curve的第一个参数是{0,1}范围内的二进制标签，第二个参数是目标分数。 你正经过probab为目标分数和y_test为二进制标签。

from sklearn.metrics import roc_curve

y_test = np.asarray([-10.54, -9.49, -9.4, -9.37, -9.36, -9.31, -9.28, -9.14, -9.11, -9.03, -9.01, -9.0, -8.99, -8.98, -8.96, -8.91, -8.9, -8.9, -8.9, -8.89, -8.88, -8.86, -8.86, -8.84, -8.83, -8.78, -8.76, -8.74, -8.74, -8.69, -8.69, -8.69, -8.67, -8.64, -8.61, -8.57, -8.51, -8.5, -8.49, -8.48, -8.4, -8.34, -8.33, -8.3, -8.29, -8.29, -8.27, -8.26, -8.25, -8.22, -8.15, -8.12, -8.1, -8.08, -8.04, -8.04, -7.96, -7.94, -7.94, -7.85, -7.83, -7.82, -7.82, -7.81, -7.76, -7.74, -7.71, -7.65, -7.57, -7.54, -7.47, -7.4, -7.39, -7.34, -7.33, -7.32, -7.27, -7.23, -7.16, -7.08, -7.05, -6.92, -6.9, -6.89, -6.86, -6.86, -6.83, -6.78, -6.73, -6.69, -6.59, -6.57, -6.4, -6.37, -6.21, -6.19, -6.16, -6.04, -6.04, -5.57, -5.54, -5.35, -5.24, -5.0, -4.92])
probas = np.asarray([1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])
fpr, tpr, thresholds = roc_curve(probas,y_test)
plt.plot(fpr, label = 'fpr')
plt.plot(tpr, label = 'tpr')
plt.legend(fontsize=16)

产量