自定义评价指标版本一(使用0.2作为阈值):
import keras.backend as K
#precision
def selfPrecision(y_true, y_pred):
true_positives = K.sum(K.cast(K.greater(K.clip(y_true * y_pred, 0, 1), 0.20), 'float32'))
pred_positives = K.sum(K.cast(K.greater(K.clip(y_pred, 0, 1), 0.20), 'float32'))
precision = true_positives / (pred_positives + K.epsilon())
return precision
#recall
def selfRecall(y_true, y_pred):
true_positives = K.sum(K.cast(K.greater(K.clip(y_true * y_pred, 0, 1), 0.20), 'float32'))
poss_positives = K.sum(K.cast(K.greater(K.clip(y_true, 0, 1), 0.20), 'float32'))
recall = true_positives / (poss_positives + K.epsilon())
return recall
#f1-measure
def selfF1Score(y_true, y_pred):
p_val = selfPrecision(y_true, y_pred)
r_val = selfRecall(y_true, y_pred)
f_val = 2*p_val*r_val / (p_val + r_val)
return f_val
函数调用:
model.compile(optimizer=rms_prop, loss=“binary_crossentropy”, metrics=[selfPrecision, selfRecall, selfF1Score])
备注:代码里面做判断的时候使用了0.2作为阈值,个人理解这个需要结合自己实际情况进行修改
自定义评价指标版本二(使用0.5作为阈值):
import keras.backend as K
#precision
def selfPrecision(y_true, y_pred):
true_positives = K.sum(K.cast(K.greater(K.clip(y_true * y_pred, 0, 1), 0.20), 'float32'))
pred_positives = K.sum(K.cast(K.greater(K.clip(y_pred, 0, 1), 0.20), 'float32'))
precision = true_positives / (pred_positives + K.epsilon())
return precision
#recall
def selfRecall(y_true, y_pred):
true_positives = K.sum(K.cast(K.greater(K.clip(y_true * y_pred, 0, 1), 0.20), 'float32'))
poss_positives = K.sum(K.cast(K.greater(K.clip(y_true, 0, 1), 0.20), 'float32'))
recall = true_positives / (poss_positives + K.epsilon())
return recall
#f1-measure
def selfF1Score(y_true, y_pred):
p_val = selfPrecision(y_true, y_pred)
r_val = selfRecall(y_true, y_pred)
f_val = 2*p_val*r_val / (p_val + r_val)
return f_val
函数调用:
model.compile(optimizer=rms_prop, loss=“binary_crossentropy”, metrics=[selfPrecision, selfRecall, selfF1Score])
备注:代码里面做判断的时候使用了0.5作为阈值,因为我的label只有0和1,预测的结果介于0和1之间,就是说只有预测值大于0.5,并且label为1的数据才算作TP
内容来源:
https://blog.csdn.net/qq_36142114/article/details/103159875?spm=1001.2014.3001.5502
自定义评价指标版本三:
from keras import backend
def rmse(y_true, y_pred):
return backend.sqrt(backend.mean(backend.square(y_pred - y_true), axis=-1))
函数调用:
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=[rmse])
内容来源:
https://www.bbsmax.com/A/6pdDVrwKJw/
留疑点:对于版本一与版本二,其区别仅仅在于判断的阈值不同,原文的作者使用0.2作为阈值,一直感觉非常奇怪,欢迎大家评论区讨论为何如此处理?