错误分类指标，无法处理二进制和连续目标的混合

Question

你好，我很困惑如何解决这个问题。这是我第一次尝试深度学习。我已经检查了其他问题的问题和答案，但仍然没有得到任何解决。如何解决这个错误

ValueError: Classification metrics can't handle a mix of binary and continuous targets

下面是我的代码:

In>>y_pred = model.predict(seq_array, batch_size=200, verbose=1)
In>>y_true = label_array
In>>print('Confusion matrixn- x-axis is true labels.n- y-axis is predicted labels')
In>>print(y_pred)
In>>print(y_true)
In>>confusion_m = confusion_matrix(y_true, y_pred)
In>>confusion_m
Out>>
20/20 [==============================] - 2s 83ms/step
Confusion matrix
- x-axis is true labels.
- y-axis is predicted labels
[[0.00791603]
[0.00798142]
[0.00804839]
...
[0.52200854]
[0.5300765 ]
[0.53883666]]
[[0.]
[0.]
[0.]
...
[1.]
[1.]
[1.]]

Answer 1

您的predict()返回分数/概率而不是类别。您应该将它们与您选择的阈值进行比较，以获得类预测(0或1)。类似confusion_matrix(y_true, (y_pred > 0.5))的东西应该可以工作(假设这些是numpy数组)。