Tensorflow計算正確率、精確率、召回率

二分類模型的評價指標

https://www.cnblogs.com/xiaoniu-666/p/10511694.html

參考tf的方法

    predictions = tf.argmax(predict, 1)
    actuals = tf.argmax(real, 1)

    ones_like_actuals = tf.ones_like(actuals)
    zeros_like_actuals = tf.zeros_like(actuals)
    ones_like_predictions = tf.ones_like(predictions)
    zeros_like_predictions = tf.zeros_like(predictions)

        Lable:      1   1   0   0
        predi:      1   0   0   1
                    Tp  Fp  Tn  Fn
    tp: = and       1
    tn = ont(or)            1

    lab-pred:       0   1   0   -1

    lab-pred>=0.6:  0   1   0   0
    fp = and(lable, lab-pred):
                    0   1   0   1

    lab-pred<=-1.0: 0   0   0   1
    not-lable:      0   0   1   1
    fn = and(not-lable, lab-pred<-1.0)

可能用到的方法：

tf.less_equal
tf.less
tf.greater_equal
tf.greater
count_nonzero

參考：

https://blog.csdn.net/sinat_35821976/article/details/81334181

https://tensorflow.google.cn/api_docs/python/tf/math/count_nonzero