def collect_train_data(begin,end):
conn,cursor = db.db_connect()
for market in config.MARKETS:
code_list = stock.get_stock_by_market(market,cursor)
for code in code_list:
collect_stock_train_data(begin,end,code,market,cursor)
conn.commit()
db.db_close(conn,cursor)
def update_daily():
global page
page = 0
conn, cursor = db.db_connect()
while page <= total_page:
print('page : ' + str(page) + ' of ' + str(total_page))
try:
updateDZJY(page, cursor)
except:
print 'traceback.print_exc():'
traceback.print_exc()
page = page + 1
continue
page = page + 1
time.sleep(2)
conn.commit()
db.db_close(conn, cursor)
def collect_price(begin, end):
conn, cursor = db.db_connect()
enddate = datetime.datetime(int(end[0:4]), int(end[5:7]), int(end[8:10]))
begindate = datetime.datetime(int(begin[0:4]), int(begin[5:7]),
int(begin[8:10]))
begin = begindate.strftime('%Y%m%d')
end = enddate.strftime('%Y%m%d')
for market in config.MARKETS:
code_list = stock.get_stock_by_market(market, cursor)
for code in code_list:
if market == 'sh':
download_price_file(begin, end, code, 0)
else:
download_price_file(begin, end, code, 1)
read_price_file(code + '.csv', market, cursor)
os.remove(code + '.csv')
conn.commit()
db.db_close(conn, cursor)
code_path = '/home/ayesha/data/models/' + code
model_path = code_path + '/' + term
module_file = tf.train.latest_checkpoint(model_path)
saver.restore(sess, module_file)
#训练
for i in range(config.DAILY_TRAINING_STEPS + 1):
for step in range(len(batch_index) - 1):
final_states, loss_ = sess.run(
[train_op, loss],
feed_dict={
X:
train_x[batch_index[step]:batch_index[step + 1]],
Y: train_y[batch_index[step]:batch_index[step + 1]]
})
#保存模型
print(
"save model : ",
saver.save(sess,
model_path + '/stock.model',
global_step=global_step))
if __name__ == '__main__':
conn, cursor = db.db_connect()
code = sys.argv[1]
batch_size = int(sys.argv[2])
time_step = int(sys.argv[3])
term = sys.argv[4]
date = sys.argv[5]
daily_train_lstm(code, batch_size, time_step, term, date, cursor)
db.db_close(conn, cursor)
def predict_lstm(code, time_step, term, date):
conn, cursor = db.db_connect()
with tf.variable_scope(code + '_' + term, reuse=None):
#输入层、输出层权重、偏置
weights = {
'in': tf.Variable(tf.random_normal([input_size, rnn_unit])),
'out': tf.Variable(tf.random_normal([rnn_unit, 1]))
}
biases = {
'in': tf.Variable(tf.constant(0.1, shape=[
rnn_unit,
])),
'out': tf.Variable(tf.constant(0.1, shape=[
1,
]))
}
X = tf.placeholder(tf.float32, shape=[None, time_step, input_size])
batch_size = tf.shape(X)[0]
time_step_tensor = tf.shape(X)[1]
w_in = weights['in']
b_in = biases['in']
input = tf.reshape(X,
[-1, input_size]) #需要将tensor转成2维进行计算，计算后的结果作为隐藏层的输入
input_rnn = tf.matmul(input, w_in) + b_in
input_rnn = tf.reshape(
input_rnn,
[-1, time_step_tensor, rnn_unit]) #将tensor转成3维，作为lstm cell的输入
cell = tf.nn.rnn_cell.BasicLSTMCell(rnn_unit)
init_state = cell.zero_state(batch_size, dtype=tf.float32)
output_rnn, final_states = tf.nn.dynamic_rnn(
cell, input_rnn, initial_state=init_state, dtype=tf.float32
) #output_rnn是记录lstm每个输出节点的结果，final_states是最后一个cell的结果
output = tf.reshape(output_rnn, [-1, rnn_unit]) #作为输出层的输入
w_out = weights['out']
b_out = biases['out']
pred = tf.matmul(output, w_out) + b_out
saver = tf.train.Saver()
mean, std, test_x, future_price = get_test_data(
code, time_step, term, date, cursor)
if len(test_x) == 0:
return
with tf.Session() as sess:
#参数恢复
code_path = '/home/ayesha/data/models/' + code
model_path = code_path + '/' + term
module_file = tf.train.latest_checkpoint(model_path)
saver.restore(sess, module_file)
test_predict = []
for step in range(len(test_x)):
prob = sess.run(pred, feed_dict={X: [test_x[step]]})
predict = prob.reshape((-1))
test_predict.extend(predict)
#预测值
test_predict = np.array(test_predict) * float(std) + float(mean)
#print(test_predict)
est_price = test_predict[-1]
print('refreshing stock est data ' + code + ' ' + date)
params = [
code, date, est_price, future_price, est_price, future_price
]
stock_est_data.refresh_stock_est_data(params, cursor)
conn.commit()
db.db_close(conn, cursor)