https://stackoverflow.com/questions/41032551/how-to-compute-receiving-operating-characteristic-roc-and-auc-in-keras
https://github.com/keras-team/keras/issues/3230#issuecomment-319208366
http://www.luozhipeng.com/?p=1225
http://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_auc_score.html
http://blog.csdn.net/zdy0_2004/article/details/44948511
keras 输出roc指标，不能每个batch输出一次，需要全部计算完再一次计算一次。使用sklearn中的metrics roc来计算。几个帖子类似
class roc_callback(keras.callbacks.Callback): def __init__(self,training_data,validation_data): self.x = training_data[0] self.y = training_data[1] self.x_val = validation_data[0] self.y_val = validation_data[1] def on_train_begin(self, logs={}): return def on_train_end(self, logs={}): return def on_epoch_begin(self, epoch, logs={}): return def on_epoch_end(self, epoch, logs={}): y_pred = self.model.predict(self.x) roc = roc_auc_score(self.y, y_pred) y_pred_val = self.model.predict(self.x_val) roc_val = roc_auc_score(self.y_val, y_pred_val) print(‘\rroc-auc: %s - roc-auc_val: %s‘ % (str(round(roc,4)),str(round(roc_val,4))),end=100*‘ ‘+‘\n‘) return def on_batch_begin(self, batch, logs={}): return def on_batch_end(self, batch, logs={}): return callbacks=[roc_callback(training_data=training_data,validation_data=validation_data)]
首先建立callbacks脚本，
my_callbacks.py如下：
import keras from sklearn.metrics import roc_auc_score import numpy as np class Histories(keras.callbacks.Callback): 6 def on_train_begin(self, logs={}): 7 self.aucs = [] 8 self.losses = [] 9 10 def on_train_end(self, logs={}): 11 return 12 13 def on_epoch_begin(self, epoch, logs={}): 14 return 15 16 def on_epoch_end(self, epoch, logs={}): 17 self.losses.append(logs.get(‘loss‘)) 18 y_pred = self.model.predict(self.validation_data[0:2]) 19 20 yp = [] 21 for i in xrange(0, len(y_pred)): 22 yp.append(y_pred[i][0]) 23 yt = [] 24 for x in self.validation_data[2]: 25 yt.append(x[0]) 26 27 auc = roc_auc_score(yt, yp) 28 self.aucs.append(auc) 29 print ‘val-loss‘,logs.get(‘loss‘), ‘ val-auc: ‘,auc, 30 print ‘\n‘ 31 32 return 33 34 def on_batch_begin(self, batch, logs={}): 35 return 36 37 def on_batch_end(self, batch, logs={}): 38 return
模型的输入为：
1 |
model = Model(inputs=[keyword1, keyword2], outputs=y) |
在每个epoch结束时计算auc并输出：
1 |
histories = my_callbacks.Histories() |
2 |
3 |
model.fit(train_x, train_y, batch_size=1024, epochs=20,shuffle=True, class_weight={1:1.0, 0:0.25}, validation_split=0.2, callbacks=[histories, model_check, lr]) |
