def eval_performance(self, model):
print 'start evaluating...'
Agent = dqn_agent_nature.dqn_agent(
gpu_id=self.gpu_id,
state_dimention=1,
enable_controller=self.enable_controller)
Agent.agent_init()
Agent.DQN.model = model
Agent.DQN.model_to_gpu()
Agent.policyFrozen = True
profit_list = []
test_profit_list = []
files = os.listdir(self.target_folder)
#train term evaluation
print 'start evaluation train term...'
for f in files:
stock_agent = env_stockmarket.Stock_agent(Agent,
self.action_split_number)
try:
traindata, trainprice = self.market.get_trainData(
f, self.input_num)
except:
continue
profit_ratio = stock_agent.trading(self.input_num, trainprice,
traindata)
profit_list.append(profit_ratio)
train_ave = np.mean(np.array(profit_list))
train_ave_Q = Agent.get_average_Q()
train_ave_reward = Agent.get_average_reward()
#test term evaluation
print 'start evaluation test term...'
for f in files:
stock_agent = env_stockmarket.Stock_agent(Agent,
self.action_split_number)
try:
traindata, trainprice = self.market.get_testData(
f, self.input_num)
except:
continue
profit_ratio = stock_agent.trading(self.input_num, trainprice,
traindata)
test_profit_list.append(profit_ratio)
test_ave = np.mean(np.array(test_profit_list))
self.train_ave_profit_list.append(train_ave)
self.test_ave_profit_list.append(test_ave)
self.ave_Q_list.append(train_ave_Q)
self.ave_reward_list.append(train_ave_reward)
print 'finish evaluation'
return train_ave, test_ave, train_ave_Q, train_ave_reward
#コントローラの設定
enable_controller = range(-args.action_split_number,
args.action_split_number + 1)
print 'enable_controller:', enable_controller
END_TRAIN_DAY = 20081230
START_TEST_DAY = 20090105
#START_TEST_DAY = 20100104
org_model = 0
#モデルの読み込み
#not use online update
if args.online_update == 0:
Agent = dqn_agent_nature.dqn_agent(gpu_id=args.gpu,
state_dimention=1,
enable_controller=enable_controller)
Agent.agent_init()
Agent.DQN.load_model(args.model)
Agent.policyFrozen = True
#use online update
elif args.online_update == 1:
Agent = dqn_agent_without_ER.dqn_agent(state_dimention=1,
enable_controller=enable_controller)
Agent.agent_init()
#オリジナルを改変しないようにコピー
with open(args.model, 'rb') as m:
print "open " + args.model
org_model = pickle.load(m)
Agent.DQN.model = copy.deepcopy(org_model)
else:
print 'make experiment folder'
os.makedirs(folder)
evaluation_freq = 100000
END_TRAIN_DAY = 20081230
START_TEST_DAY = 20090105
n_epoch = 2000
agent_state = 4
#コントローラの設定
enable_controller = range( - args.action_split_number,args.action_split_number + 1)
print 'enable_controller:',enable_controller
start_time = time.clock()
Agent = dqn_agent_nature.dqn_agent(gpu_id=args.gpu,enable_controller = enable_controller,state_dimention=args.input_num * args.channel + agent_state,batchsize=args.batchsize,historysize=args.historysize,epsilon_discount_size=args.epsilon_discount_size,arch=args.arch)
Agent.agent_init()
market = env_stockmarket.StockMarket(END_TRAIN_DAY,START_TEST_DAY,u_vol=u_vol,u_ema=u_ema,u_rsi=u_rsi,u_macd=u_macd,u_stoch=u_stoch,u_wil=u_wil)
evaluater = evaluation_performance.Evaluation(args.gpu,market,eval_folder,folder,args.input_num,args.action_split_number,args.arch)
print 'epoch:', n_epoch
with open(folder + 'settings.txt', 'wb') as o:
o.write('epoch:' + str(n_epoch) + '\n')
o.write('data_folder:' + str(args.data_folder) + '\n')
o.write('input:' + str(args.input_num) + '\n')
if os.path.isdir(folder) == True:
print 'this experiment name is existed'
print 'please change experiment name'
raw_input()
else:
print 'make experiment folder'
os.makedirs(folder)
END_TRAIN_DAY = 20081230
START_TEST_DAY = 20090105
n_epoch = 1000
start_time = time.clock()
Agent = dqn_agent_nature.dqn_agent(gpu_id=args.gpu,state_dimention=args.input_num * args.channel + 2,batchsize=args.batchsize,historysize=args.historysize,epsilon_discount_size=args.epsilon_discount_size,targetFlag = targetFlag)
Agent.agent_init()
market = env_stockmarket.StockMarket(END_TRAIN_DAY,START_TEST_DAY,u_vol=u_vol,u_ema=u_ema,u_rsi=u_rsi,u_macd=u_macd,u_stoch=u_stoch,u_wil=u_wil)
evaluater = evaluation_performance.Evaluation(args.gpu,market,args.data_folder,folder,args.input_num)
print 'epoch:', n_epoch
with open(folder + 'settings.txt', 'wb') as o:
o.write('epoch:' + str(n_epoch) + '\n')
o.write('data_folder:' + str(args.data_folder) + '\n')
o.write('input:' + str(args.input_num) + '\n')
o.write('channel:' + str(args.channel) + '\n')
folder = './test_result/' + args.experiment_name + '/'
if os.path.isdir(folder) == True:
print 'this experiment name is existed'
print 'please change experiment name'
raw_input()
else:
print 'make experiment folder'
os.makedirs(folder)
END_TRAIN_DAY = 20081230
#START_TEST_DAY = 20090105
START_TEST_DAY = 20100104
#モデルの読み込み
Agent = dqn_agent_nature.dqn_agent()
Agent.agent_init()
Agent.DQN.load_model(args.model)
Agent.policyFrozen = True
market = env_stockmarket.StockMarket(END_TRAIN_DAY,START_TEST_DAY,u_vol=u_vol,u_ema=u_ema,u_rsi=u_rsi,u_macd=u_macd,u_stoch=u_stoch,u_wil=u_wil)
files = os.listdir("./nikkei100")
Agent.init_max_Q_list()
Agent.init_reward_list()
profit_list = []
for f in files:
print f