def load_model(self, model_name):
self._model_name = model_name
model_path = os.path.join(get_parent_dir(), 'data', 'models',
model_name)
self.sess = tf.Session()
saver = tf.train.Saver()
saver.restore(sess=self.sess, save_path=model_path)
def save_model(self, model_name):
self._model_name = model_name
model_path = os.path.join(get_parent_dir(), 'data', 'models',
model_name)
saver = tf.train.Saver()
save_path = saver.save(self.sess, model_path)
logger.info("Saved to path:{0}".format(save_path))
def load_model(self, model_name):
model_path = os.path.join(get_parent_dir(), 'data', 'models',
'{0}'.format(model_name))
try:
self.model = joblib.load(model_path)
except Exception as ex:
logger.info('Fail to load model: {0} with error:{1}'.format(
model_path, ex))
return self.model
def init(context):
# model_path = os.path.join(get_parent_dir(), 'data', 'models', 'stock_selection_{0}'.format(model_name))
model_path = os.path.join(get_parent_dir(), 'data', 'models',
'ridge_20150103_20181231_0.9_000905.ZICN')
feature_names = get_selected_features()
context.features = load_cache_features(__config__['base']['start_date'],
__config__['base']['end_date'],
__config__['base']['benchmark'])
context.model = joblib.load(model_path)
context.feature_names = feature_names
context.neu = False
def train_model(self,
train_X,
train_Y,
acc,
n_epochs=100,
batch_size=50,
model_name=None):
'''
:param train_X: feature input,N*K vector; N is sec number, K is the total number of sub-type feature
:param train_Y: trained label; N*1 vector; e.g. the return of stock
:param acc: industry feature input, N*I; N is sec number
:param n_epochs: trained epoch
:param batch_size: trained batch_size
:param model_name:
:return:
'''
init = tf.global_variables_initializer()
# saver = tf.train.Saver()
self._dataset = DataSet(train_X, train_Y, acc)
with tf.Session() as self.sess:
train_writer = tf.summary.FileWriter(
"E:\pycharm\quant_geek\quant_models\data\models",
self.sess.graph)
init.run()
for epoch in range(n_epochs):
logger.info('Run the {0} epoch out of {1}, with '.format(
epoch, n_epochs))
for iteration in range(self._dataset.num_examples //
batch_size):
x_batch, y_batch, acc = self._dataset.next_batch(
batch_size)
self.sess.run(
[self.training_op],
feed_dict={
self.feature_inputs: x_batch,
self.train_Y: y_batch,
self.indust_inputs: acc
})
x_test, y_test, acc_test = self._dataset.next_batch(batch_size)
test_loss, test_summary = self.sess.run(
[self.loss, self.summary_op],
feed_dict={
self.feature_inputs: x_test,
self.train_Y: y_test,
self.indust_inputs: acc_test
})
self._test_loss.append(test_loss)
logger.info('epoch: {0}, test_loss:{1}'.format(
epoch, test_loss))
train_writer.add_summary(test_summary, epoch)
if model_name:
self._model_name = model_name
model_path = os.path.join(get_parent_dir(), 'data', 'models',
model_name)
saver = tf.train.Saver()
save_path = saver.save(self.sess, model_path)
logger.info("Saved to path:{0}".format(save_path))
self._test_loss = np.array(self._test_loss)
logger.info('mean for test loss:{0}, std:{1}, var:{2}'.format(
self._test_loss.mean(), self._test_loss.std(),
self._test_loss.var()))
def save_model(self, model_path):
model_path = os.path.join(get_parent_dir(), 'data', 'models',
model_path)
joblib.dump(self.model, model_path, protocol=2)