【宽客策略源码】多因子选股(股票)

geek168 数据策略 2019-8-11 15:49 52004人围观

本策略每隔1个月定时触发,根据Fama-French三因子模型对每只股票进行回归，得到其alpha值。假设Fama-French三因子模型可以完全解释市场，则alpha为负表明市场低估该股，因此应该买入。策略思路：

计算市场收益率、个股的账面市值比和市值,并对后两个进行了分类,
根据分类得到的组合分别计算其市值加权收益率、SMB和HML.
对各个股票进行回归(假设无风险收益率等于0)得到alpha值.
选取alpha值小于0并为最小的10只股票进入标的池
平掉不在标的池的股票并等权买入在标的池的股票

回测数据:SHSE.000300的成份股回测时间:2017-07-01 08:00:00到2017-10-01 16:00:0

# coding=utf-8
from __future__ import print_function, absolute_import, unicode_literals
import numpy as np
from gm.api import *
from pandas import DataFrame
'''
'''
def init(context):
# 每月第一个交易日的09:40 定时执行algo任务
schedule(schedule_func=algo, date_rule='1m', time_rule='09:40:00')
print(order_target_percent(symbol='SHSE.600000', percent=0.5, order_type=OrderType_Market,
position_side=PositionSide_Long))
# 数据滑窗
context.date = 20
# 设置开仓的最大资金量
context.ratio = 0.8
# 账面市值比的大/中/小分类
context.BM_BIG = 3.0
context.BM_MID = 2.0
context.BM_SMA = 1.0
# 市值大/小分类
context.MV_BIG = 2.0
context.MV_SMA = 1.0
# 计算市值加权的收益率,MV为市值的分类,BM为账目市值比的分类
def market_value_weighted(stocks, MV, BM):
select = stocks[(stocks.NEGOTIABLEMV == MV) & (stocks.BM == BM)]
market_value = select['mv'].values
mv_total = np.sum(market_value)
mv_weighted = [mv / mv_total for mv in market_value]
stock_return = select['return'].values
# 返回市值加权的收益率的和
return_total = []
for i in range(len(mv_weighted)):
return_total.append(mv_weighted[i] * stock_return[i])
return_total = np.sum(return_total)
return return_total
def algo(context):
# 获取上一个交易日的日期
last_day = get_previous_trading_date(exchange='SHSE', date=context.now)
# 获取沪深300成份股
context.stock300 = get_history_constituents(index='SHSE.000300', start_date=last_day,
end_date=last_day)[0]['constituents'].keys()
# 获取当天有交易的股票
not_suspended = get_history_instruments(symbols=context.stock300, start_date=last_day, end_date=last_day)
not_suspended = [item['symbol'] for item in not_suspended if not item['is_suspended']]
fin = get_fundamentals(table='tq_sk_finindic', symbols=not_suspended, start_date=last_day, end_date=last_day,
fields='PB,NEGOTIABLEMV', df=True)
# 计算账面市值比,为P/B的倒数
fin['PB'] = (fin['PB'] ** -1)
# 计算市值的50%的分位点,用于后面的分类
size_gate = fin['NEGOTIABLEMV'].quantile(0.50)
# 计算账面市值比的30%和70%分位点,用于后面的分类
bm_gate = [fin['PB'].quantile(0.30), fin['PB'].quantile(0.70)]
fin.index = fin.symbol
x_return = []
# 对未停牌的股票进行处理
for symbol in not_suspended:
# 计算收益率
close = history_n(symbol=symbol, frequency='1d', count=context.date + 1, end_time=last_day, fields='close',
skip_suspended=True, fill_missing='Last', adjust=ADJUST_PREV, df=True)['close'].values
stock_return = close[-1] / close[0] - 1
pb = fin['PB'][symbol]
market_value = fin['NEGOTIABLEMV'][symbol]
# 获取[股票代码. 股票收益率, 账面市值比的分类, 市值的分类, 流通市值]
if pb < bm_gate[0]:
if market_value < size_gate:
label = [symbol, stock_return, context.BM_SMA, context.MV_SMA, market_value]
else:
label = [symbol, stock_return, context.BM_SMA, context.MV_BIG, market_value]
elif pb < bm_gate[1]:
if market_value < size_gate:
label = [symbol, stock_return, context.BM_MID, context.MV_SMA, market_value]
else:
label = [symbol, stock_return, context.BM_MID, context.MV_BIG, market_value]
elif market_value < size_gate:
label = [symbol, stock_return, context.BM_BIG, context.MV_SMA, market_value]
else:
label = [symbol, stock_return, context.BM_BIG, context.MV_BIG, market_value]
if len(x_return) == 0:
x_return = label
else:
x_return = np.vstack([x_return, label])
stocks = DataFrame(data=x_return, columns=['symbol', 'return', 'BM', 'NEGOTIABLEMV', 'mv'])
stocks.index = stocks.symbol
columns = ['return', 'BM', 'NEGOTIABLEMV', 'mv']
for column in columns:
stocks[column] = stocks[column].astype(np.float64)
# 计算SMB.HML和市场收益率
# 获取小市值组合的市值加权组合收益率
smb_s = (market_value_weighted(stocks, context.MV_SMA, context.BM_SMA) +
market_value_weighted(stocks, context.MV_SMA, context.BM_MID) +
market_value_weighted(stocks, context.MV_SMA, context.BM_BIG)) / 3
# 获取大市值组合的市值加权组合收益率
smb_b = (market_value_weighted(stocks, context.MV_BIG, context.BM_SMA) +
market_value_weighted(stocks, context.MV_BIG, context.BM_MID) +
market_value_weighted(stocks, context.MV_BIG, context.BM_BIG)) / 3
smb = smb_s - smb_b
# 获取大账面市值比组合的市值加权组合收益率
hml_b = (market_value_weighted(stocks, context.MV_SMA, 3) +
market_value_weighted(stocks, context.MV_BIG, context.BM_BIG)) / 2
# 获取小账面市值比组合的市值加权组合收益率
hml_s = (market_value_weighted(stocks, context.MV_SMA, context.BM_SMA) +
market_value_weighted(stocks, context.MV_BIG, context.BM_SMA)) / 2
hml = hml_b - hml_s
close = history_n(symbol='SHSE.000300', frequency='1d', count=context.date + 1,
end_time=last_day, fields='close', skip_suspended=True,
fill_missing='Last', adjust=ADJUST_PREV, df=True)['close'].values
market_return = close[-1] / close[0] - 1
coff_pool = []
# 对每只股票进行回归获取其alpha值
for stock in stocks.index:
x_value = np.array([[market_return], [smb], [hml], [1.0]])
y_value = np.array([stocks['return'][stock]])
# OLS估计系数
coff = np.linalg.lstsq(x_value.T, y_value)[0][3]
coff_pool.append(coff)
# 获取alpha最小并且小于0的10只的股票进行操作(若少于10只则全部买入)
stocks['alpha'] = coff_pool
stocks = stocks[stocks.alpha < 0].sort_values(by='alpha').head(10)
symbols_pool = stocks.index.tolist()
positions = context.account().positions()
# 平不在标的池的股票
for position in positions:
symbol = position['symbol']
if symbol not in symbols_pool:
order_target_percent(symbol=symbol, percent=0, order_type=OrderType_Market,
position_side=PositionSide_Long)
print('市价单平不在标的池的', symbol)
# 获取股票的权重
percent = context.ratio / len(symbols_pool)
# 买在标的池中的股票
for symbol in symbols_pool:
order_target_percent(symbol=symbol, percent=percent, order_type=OrderType_Market,
position_side=PositionSide_Long)
print(symbol, '以市价单调多仓到仓位', percent)
if __name__ == '__main__':
'''
strategy_id策略ID,由系统生成
filename文件名,请与本文件名保持一致
mode实时模式:MODE_LIVE回测模式:MODE_BACKTEST
token绑定计算机的ID,可在系统设置-密钥管理中生成
backtest_start_time回测开始时间
backtest_end_time回测结束时间
backtest_adjust股票复权方式不复权:ADJUST_NONE前复权:ADJUST_PREV后复权:ADJUST_POST
backtest_initial_cash回测初始资金
backtest_commission_ratio回测佣金比例
backtest_slippage_ratio回测滑点比例
'''
run(strategy_id='strategy_id',
filename='main.py',
mode=MODE_BACKTEST,
token='token_id',
backtest_start_time='2017-07-01 08:00:00',
backtest_end_time='2017-10-01 16:00:00',
backtest_adjust=ADJUST_PREV,
backtest_initial_cash=10000000,
backtest_commission_ratio=0.0001,
backtest_slippage_ratio=0.0001)