model.fit(
train_X,
train_y,
epochs=epochs,
# callbacks=[early_stopping, checkpoint],
batch_size=batch_size,
validation_split=0.1,
validation_data=(val_X, val_y),
# class_weight={0:1, 1:3}, # 样本均衡
)
print('test AUC: %f' % model.evaluate(test_X, test_y)[1])
# ------------- model evaluation in test dataset ----
train_predictions_weighted = model.predict(train_X, batch_size=batch_size)
test_predictions_weighted = model.predict(test_X, batch_size=batch_size)
# ------------- confusion matrix
from sklearn.metrics import confusion_matrix, roc_curve
import matplotlib.pyplot as plt
import seaborn as sns
def plot_cm(labels, predictions, p=0.5):
cm = confusion_matrix(labels, predictions > p)
plt.figure(figsize=(5, 5))
sns.heatmap(cm, annot=True, fmt="d")
plt.title('Confusion matrix @{:.2f}'.format(p))
plt.ylabel('Actual label')
plt.xlabel('Predicted label')
deep_input.append(wide_features.values)
my_callbacks = [
EarlyStopping(monitor='loss',
min_delta=1e-5,
patience=3,
verbose=1,
mode='min')
]
model.fit(x=deep_input,
y=final_tags,
batch_size=64,
epochs=100,
verbose=1,
callbacks=my_callbacks)
pred = model.predict(deep_input, batch_size=2**14)
pred_real = target_mms.inverse_transform(pred)
s0, s1 = 0, 0
for i, x in enumerate(pred_real):
if pred[i] > 0:
s1 += 1
print(deep_features.iloc[i, 0], deep_features.iloc[i, -1],
final_tags[i], pred[i])
else:
s0 += 1
print('s0:', s0, 's1:', s1)