def make_env():
# stock_db
stock_db = StockDatabase(db_path)
# sampler
#ticker_names_sampler = TickerSampler(all_ticker_names=ticker_codes,
# sampling_ticker_number=ticker_number)
ticker_names_sampler = ConstSamper(
TickerSampler(ticker_codes, ticker_number).sample()) # 固定する
start_datetime_sampler = DatetimeSampler(start_datetime=start_datetime,
end_datetime=end_datetime,
episode_length=episode_length,
freq_str=freq_str)
portfolio_sampler = PortfolioVectorSampler(ticker_number + 1)
sampler_manager = SamplerManager(
ticker_names_sampler=ticker_names_sampler,
datetime_sampler=start_datetime_sampler,
portfolio_vector_sampler=portfolio_sampler,
)
# PriceSupplierの設定
price_supplier = StockDBPriceSupplier(
stock_db,
[], # 最初は何の銘柄コードも指定しない
episode_length,
freq_str,
interpolate=True)
# PortfolioTransformerの設定
portfolio_transformer = PortfolioTransformer(
price_supplier,
portfolio_restrictor=PortfolioRestrictorIdentity(),
use_ohlc="Close",
initial_all_assets=1e6, # 学習には関係ない
fee_calculator=FeeCalculatorFree())
# TradeEnvの設定
trade_env = TradeEnv(portfolio_transformer,
sampler_manager,
window=window,
fee_const=0.0025)
return trade_env
if __name__ == "__main__":
import argparse
print("[{}] schedule program start".format(str(datetime.datetime.now())))
parser = argparse.ArgumentParser(
description='insert data to database with scheduling')
parser.add_argument("--tempfile",
action="store_true",
help="tempfileを利用するかどうか")
args = parser.parse_args()
# 必要なインスタンス
db_path = Path("db/stock_db") / Path("stock.db")
stock_db = StockDatabase(db_path,
column_upper_limit=1000,
table_name_base="table")
nikkei_code_file_path = Path("get_stock_price") / Path("nikkei225.csv")
tosho_code_file_path = Path("get_stock_price") / Path("tosho.csv")
stock_loader = YahooFinanceStockLoaderMin(None,
past_day=5,
stop_time_span=2.0,
is_use_stop=False) #ストップしない
nikkei_kobetsu_insert = CsvKobetsuInsert(nikkei_code_file_path,
stock_loader,
stock_db,
stock_group="nikkei_255",
use_tempfile=args.tempfile)
tosho_kobetsu_insert = CsvKobetsuInsert(tosho_code_file_path,
import torch.nn.functional as F
import pandas as pd
import collections
import datetime
from pytz import timezone
from pathlib import Path
from get_stock_price import StockDatabase
from envs_ver2 import OneStockEnv, NormalizeState, NormalizeReward
db_path = Path("E:/システムトレード入門/trade_system_git_workspace/db/stock_db"
) / Path("stock.db")
stock_db = StockDatabase(db_path)
jst_timezone = timezone("Asia/Tokyo")
start_datetime = jst_timezone.localize(
datetime.datetime(2020, 11, 1, 0, 0, 0))
end_datetime = jst_timezone.localize(
datetime.datetime(2020, 12, 1, 0, 0, 0))
#end_datetime = get_next_workday_jp(start_datetime, days=11) # 営業日で一週間(5日間)
#stock_names = "4755"
stock_names = "9984"
#stock_names = ["6502","4755"]
#stock_list = ["4755","9984","6701","7203","7267"]
use_ohlc = "Close"
if __name__ == "__main__":
import sys
sys.path.append(r"E:\システムトレード入門\tutorials\rl\pfrl")
sys.path.append(r"E:\システムトレード入門\trade_system_git_workspace")
import datetime
from pytz import timezone
from pathlib import Path
import pfrl
from get_stock_price import StockDatabase
from envs_ver2 import OneStockEnv, NormalizeState, NormalizeReward
db_path = Path("E:/システムトレード入門/trade_system_git_workspace/db/stock_db"
) / Path("stock.db")
stock_db = StockDatabase(db_path)
jst_timezone = timezone("Asia/Tokyo")
start_datetime = jst_timezone.localize(
datetime.datetime(2020, 11, 1, 0, 0, 0))
end_datetime = jst_timezone.localize(
datetime.datetime(2020, 12, 1, 0, 0, 0))
#end_datetime = get_next_workday_jp(start_datetime, days=11) # 営業日で一週間(5日間)
#stock_names = "4755"
#stock_names = "9984"
stock_names = "6502"
#stock_names = ["6502","4755"]
#stock_list = ["4755","9984","6701","7203","7267"]
stock_df = stock_db.search_span(stock_names=stock_names,
def make_env(
db_path,
csv_path,
is_ticker_sample=True,
start_datetime=jst.localize(datetime.datetime(2020, 11, 10, 0, 0, 0)),
end_datetime=jst.localize(datetime.datetime(2020, 11, 20, 0, 0, 0)),
episode_length=300,
window=np.arange(0, 50),
ticker_number=19,
fee_const=0.0025,
):
ticker_codes_df = pd.read_csv(csv_path, header=0) # 自分で作成
ticker_codes = ticker_codes_df["code"].values.astype(str).tolist()
# stock_db
stock_db = StockDatabase(db_path)
# sampler
if is_ticker_sample:
ticker_names_sampler = TickerSampler(
all_ticker_names=ticker_codes,
sampling_ticker_number=ticker_number)
else:
ticker_names_sampler = ConstSamper(
TickerSampler(ticker_codes, ticker_number).sample()) # 固定する
start_datetime_sampler = DatetimeSampler(start_datetime=start_datetime,
end_datetime=end_datetime,
episode_length=episode_length,
freq_str=freq_str)
portfolio_sampler = PortfolioVectorSampler(ticker_number + 1)
sampler_manager = SamplerManager(
ticker_names_sampler=ticker_names_sampler,
datetime_sampler=start_datetime_sampler,
portfolio_vector_sampler=portfolio_sampler,
)
# PriceSupplierの設定
price_supplier = StockDBPriceSupplier(
stock_db,
[], # 最初は何の銘柄コードも指定しない
episode_length,
freq_str,
interpolate=True)
# PortfolioTransformerの設定
portfolio_transformer = PortfolioTransformer(
price_supplier,
portfolio_restrictor=PortfolioRestrictorIdentity(),
use_ohlc="Close",
initial_all_assets=1e6, # 学習には関係ない
fee_calculator=FeeCalculatorFree())
# TradeEnvの設定
trade_env = TradeEnv(portfolio_transformer,
sampler_manager,
window=window,
fee_const=fee_const)
return trade_env
use_x_range=use_x_range,
use_y_range=use_y_range,
data_left_times=data_left_times,
is_notebook=is_notebook,
use_formatter=use_formatter)
if __name__ == "__main__":
from tornado.ioloop import IOLoop # サーバーをたてるのに必要
from bokeh.server.server import Server # サーバーを立てるのに必要
from pytz import timezone
from get_stock_price import StockDatabase
db_path = Path("db/stock_db") / Path("stock.db")
stock_db = StockDatabase(db_path)
stock_name = "4755" # 楽天
stock_timestamp_df = stock_db.stock_timestamp(stock_names=["4755"],
to_tokyo=True)
day_before = stock_timestamp_df.loc[0,
"min_datetime"] + datetime.timedelta(
days=10) # 最初の日時から次の日とする．
# 日時の取得
jst_timezone = timezone("Asia/Tokyo")
start_time = jst_timezone.localize(
datetime.datetime(day_before.year, day_before.month, day_before.day, 9,
0, 0))
#start_time = jst_timezone.localize(datetime.datetime(day_before.year, day_before.month, day_before.day, 12, 30, 0))
end_time = jst_timezone.localize(