def run_stock():
n_epoch = 100
mean = []
for x in range(n_epoch):
env = Env(STOCK.Baidu)
env.set_count(999) # from 1000 to 2000
for i in range(1000):
action = random.random() * 2 - 1 # [-1, 1]
env.step(action)
mean.append(env.asset - 10000)
# while True:
# action = random.random() * 2 - 1 # [-1, 1]
# observation_, reward, done = env.step(action)
#
# if done:
# mean.append(env.asset - 10000)
# break
print(np.mean(mean), np.var(mean))
# end of game
print('game over')
from stock_env_discrete import StockEnv as StockEnvD
from stock_env_discrete import STOCK as STOCKD
from sklearn.externals import joblib
import warnings
if __name__ == '__main__':
warnings.filterwarnings("ignore") # ignore warnings
stockHMM = StockHMM(STOCK.Baidu)
# load model
stockHMM.model = joblib.load("BaiDuHMM.pkl")
print('Continuous: ')
env = StockEnv(STOCK.Baidu)
# [1000, 2000)
env.set_count(999)
for x in range(1000):
p_states = stockHMM.predict(x + 1000)
# order: 2 3 4 1 0
my_action = p_states[
2] + p_states[3] * 0.5 - p_states[1] * 0.5 - p_states[0]
env.step(my_action)
print(env.asset - 10000)
print('Discrete: ')
env = StockEnvD(STOCKD.Baidu)
# [1000, 2000)
env.set_count(999)
for x in range(1000):
p_states = stockHMM.predict(x + 1000)