In [ ]:

    

import matplotlib as mpl
mpl.use('tkagg')
import numpy as np
import pandas as pd
import talib
from logging import getLogger, DEBUG, INFO, WARN, ERROR, CRITICAL
import os
import logging
from logging import StreamHandler, LogRecord
from episode_logger import EpisodeLogger
from test_output_logger import TestOutputLogger

from hist_data import HistData, BitcoinHistData
from fx_trade import FXTrade
from bitcoin_trade import BitcoinTrade
from deep_fx import DeepFX
from debug_tools import DebugTools


    

In [ ]:

    

import crcmod
class LogRecordWithHexThereadID(logging.LogRecord):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.hex_threadid = self._calc_hex(self.process)

    def _calc_hex(self, digit_value):
        return hex(digit_value)

def init_logger(sd_loglevel=logging.WARN, stream_loglevel=logging.CRITICAL):
    logging.setLogRecordFactory(LogRecordWithHexThereadID)
    logger = logging.getLogger('deepfx')
    logger.setLevel(sd_loglevel)
    formatter = logging.Formatter('[%(hex_threadid)s] %(message)s')

    if sd_loglevel:
        import google
        from google.cloud.logging import Client
        from google.cloud.logging.handlers import CloudLoggingHandler
        client = google.cloud.logging.Client \
            .from_service_account_json(os.environ.get('GOOGLE_SERVICE_ACCOUNT_JSON_PATH'))
        handler = CloudLoggingHandler(client, name='deepfx')
        handler.setLevel(sd_loglevel)
        handler.setFormatter(formatter)
        logger.addHandler(handler)
        handler = None

    if stream_loglevel:
        handler = StreamHandler()
        handler.setLevel(stream_loglevel)
        handler.setFormatter(formatter)
        logger.addHandler(handler)
        handler = None

    return logger


    

In [ ]:

    

deepfx_logger = init_logger(stream_loglevel=None)
deepfx_logger.critical('DeepFX Started: %s' % DebugTools.now_str())
deepfx_logger.debug   ('loglevel debug    test')
deepfx_logger.info    ('loglevel info     test')
deepfx_logger.warning ('loglevel warn     test')
deepfx_logger.error   ('loglevel error    test')
deepfx_logger.critical('loglevel critical test')


    

In [ ]:

    

is_for_fx = False
is_for_bitcoin = True
is_to_train = True


    

In [ ]:

    

if is_for_fx:
    hd = HistData(csv_path = 'historical_data/DAT_ASCII_USDJPY_M1_201710_h1.csv',
                     begin_date='2017-10-02T00:00:00',
                     end_date='2017-10-02T01:59:59')
elif is_for_bitcoin:
    csv_path = 'historical_data/coincheckJPY_1-min_data_2014-10-31_to_2017-10-20_h1.csv'
    if is_to_train:
        begin_date='2017-09-01T00:00:00'
        end_date='2017-09-30T23:59:59'
    else:
        begin_date='2017-10-01T00:00:00'
        end_date='2017-10-10T23:59:59'
        
    hd = HistData(csv_path, begin_date, end_date)


    

In [ ]:

    

hd.data()
len(hd.data())


    

In [ ]:

    

if is_for_fx:
    env = FXTrade(1000000, 0.08, hd, logger=deepfx_logger)
    #env = FXTrade(1000000, 0.08, h, logger=logger)
    prepared_model_filename = None #'Keras-RL_DQN_FX_model_meanq1.440944e+06_episode00003.h5'
    dfx = DeepFX(env, prepared_model_filename=prepared_model_filename, steps = 1000000, logger=deepfx_logger)
elif is_for_bitcoin:
    env = BitcoinTrade(10000000, None, hd, logger=deepfx_logger, amount_unit=0.001)
    #env = FXTrade(1000000, 0.08, h, logger=logger)
    prepared_model_filename = None #'Keras-RL_DQN_FX_model_meanq2.149058e+07_episode05184.h5'
    dfx = DeepFX(env, prepared_model_filename=prepared_model_filename, steps = 10000000, logger=deepfx_logger)
    #dfx = DeepFX(env, prepared_model_filename=prepared_model_filename, steps = 1000, logger=deepfx_logger)


    

In [ ]:

    

if is_to_train:
    dfx.train(is_for_time_measurement=True)
else:
    dfx.test([TestOutputLogger(hd)])


    

In [ ]:

    

deepfx_logger.critical('DeepFX Finished: %s' % DebugTools.now_str())


    

In [ ]:

    

import os
import subprocess
if os.environ.get('SLACK_WEBHOOK_URL') and os.environ.get('GOOGLE_STACKDRIVER_URL'):
    google_stackdriver_url = os.environ.get('GOOGLE_STACKDRIVER_URL')
    payload = '{"username":"deepfx","icon_emoji":":+1:","channel":"deepfx","attachments":[{"color":"#36a64f","title":"DeepFX Finished","title_link":"%s","text":"<@%s> DeepFX Finished"}]}' % (google_stackdriver_url, os.environ.get('SLACK_NOTIFY_RECIEVE_USER'))
    command = ['curl']
    command.append('-XPOST')
    command.append('-HContent-Type: application/json')
    command.append("-d%s" % payload)
    command.append(os.environ.get('SLACK_WEBHOOK_URL'))
    subprocess.run(command)
else:
    print('Skipped Slack Notification.')


    

In [ ]:

    

%matplotlib notebook
import matplotlib.pyplot as plt
import numpy as np
data = hd.data()['Close']
x = data.index
y = data.values
sd = 1
upper, middle, lower = talib.BBANDS(data.values, timeperiod=20, matype=talib.MA_Type.SMA, nbdevup=sd, nbdevdn=sd)
[plt.plot(x, val) for val in [y, upper, middle, lower]]


    

In [ ]:

    

data.values