def fit_predict(self, sess, input_data):
total_loss = []
collect_time = []
total_pred = []
total_labels = []
step = 0
try:
for batch_data in input_data:
batch_x, batch_sNum, batch_wNum, batch_labels = batch_data[
0], batch_data[1], batch_data[2], batch_data[3]
step += 1
total_labels.extend(batch_labels)
feed_dict = {
self.model.ph_input: batch_x,
self.model.ph_labels: batch_labels,
self.model.ph_sNum: batch_sNum,
self.model.ph_wNum: batch_wNum,
self.model.ph_train: False,
}
start_stamp = time.time()
global_step, summary, loss, y_pred = sess.run(
[
self.model.global_step, self.merged, self.model.loss,
self.model.prediction
],
feed_dict=feed_dict)
self.test_writer.add_summary(summary, step + global_step)
self.test_writer.flush()
end_stamp = time.time()
time_stamp = end_stamp - start_stamp
collect_time.append(time_stamp)
total_loss.append(loss)
total_pred.extend(y_pred)
except:
traceback.print_exc()
exit()
total_labels = np.argmax(total_labels, axis=1)
correct_predictions = float(sum(total_pred == total_labels))
accuracy = correct_predictions / float(len(total_labels))
pre, recall, f1 = score(total_pred, total_labels)
avg_loss = np.mean(total_loss)
return avg_loss, accuracy, pre, recall, f1
def classifier(train_X, train_y, test_X, test_y):
# train_X, train_y, test_X, test_y = np.array(train_X), np.array(train_y), np.array(test_X), np.array(test_y)
print(np.shape(train_X))
# clf = OneVsRestClassifier(SVC(kernel='linear'))
clf = svm.SVC()
# clf = LogisticRegression()
clf.fit(train_X, train_y)
predict_y = clf.predict(test_X)
print(predict_y)
# precision = precision_score(test_y,predict_y)
acc = metric_acc(test_y, predict_y)
pre, recall, f1 = score(test_y, predict_y)
print("acc:", acc)
print("pre:", pre)
print("recall:", recall)
print("f1:", f1)
return acc, recall, f1