config.num_steps,
shuffle=False):
data_feed = {
inputs: batch_X,
targets: batch_Y,
learning_rate: 1.0,
keep_prob: 1.0
}
predict = sess.run([prediction], data_feed)
predict = predict[0]
final_predict = np.append(final_predict, predict)
final_predict = np.array(final_predict).reshape(
len(final_predict), 1)
print("0.5 threshold")
classified = [utils.get_class(x, 0.5) for x in final_predict]
f, _ = utils.get_score_and_confusion_matrix(labels, classified)
print("Best threshold")
best_threshold = utils.find_best_threshold(
labels, final_predict)
best_classified = [
utils.get_class(x, best_threshold) for x in final_predict
]
best_f, _ = utils.get_score_and_confusion_matrix(
labels, best_classified)
if which_data == "training":
train_score.append(round(sp.average(best_f) * 100, 2))
else:
validation_score.append(round(sp.average(best_f) * 100, 2))
classified = [
config.batch_size,
config.num_steps,
shuffle=False):
data_feed = {inputs: batch_X, keep_prob: 1.0}
predict = sess.run([prediction], data_feed)
predict = predict[0]
final_predict = np.append(final_predict, predict)
final_predict = np.array(final_predict).reshape(
len(final_predict), 1)
# print("0.5 threshold")
# classified = [utils.get_class(x, 0.5) for x in final_predict]
# f, _ = utils.get_score_and_confusion_matrix(labels, classified)
# print("Best threshold")
# best_threshold = utils.find_best_threshold(labels, final_predict)
# best_classified = [utils.get_class(x, best_threshold) for x in final_predict]
best_f, _ = utils.get_score_and_confusion_matrix(
labels, final_predict)
if which_data == "training":
train_score.append(round(sp.average(best_f) * 100, 2))
else:
validation_score.append(round(sp.average(best_f) * 100, 2))
# classified = [utils.get_class(x, best_threshold) for x in final_predict]
# # Check in all the cases we correctly predict label is explosive 1, how many have the current labels 1
# current_check = []
# current_labels = reader.current_labels.values
# for index, value in enumerate(classified):
# if value + labels[index] == 2: # correctly recognition 1
# current = current_labels[index]
# if current >= 1:
# current_check.append(1)
