def ShiftTradingDay(date,shift=5):
# 获取所有的交易日，返回一个包含所有交易日的 list,元素值为 datetime.date 类型.
tradingday = jqdata.get_all_trade_days()
# 得到date之后shift天那一天在列表中的行标号 返回一个数
shiftday_index = list(tradingday).index(date)+shift
# 根据行号返回该日日期 为datetime.date类型
return tradingday[shiftday_index]
def get_bonds_history_factors(self, interval=1):
"""
获取任意指数一段时间的历史转债指数算术平均价格、转债指数溢价率 估值列表，通过计算当前的估值在历史估值的百分位，来判断当前市场的估值高低。
由于加权方式可能不同，可能各个指数公开的估值数据有差异，但用于判断估值相对高低没有问题
input：
interval: 计算指数估值的间隔天数，增加间隔时间可提高计算性能
output：
result: 指数历史估值的 DataFrame，index 为时间，列为市场总量，低估转债总量，平均价格， 溢价率
"""
all_days = get_all_trade_days()
total_markets = []
underrate_markets = []
avg_prices = []
avg_premium_ratios = []
days = []
begin = datetime.strptime(self._begin_date, '%Y-%m-%d').date()
end = datetime.strptime(self._end_date, '%Y-%m-%d').date()
i = 0
for day in all_days:
if (day < begin or day > end):
continue
i += 1
if (i % interval != 0):
continue
logging.info("======统计历史:{}================".format(
day.strftime("%Y-%m-%d")))
bond_list = self.get_bonds(day.strftime("%Y-%m-%d"))
total_market, underrate_market, avg_price, avg_premium_ratio = self.get_bonds_factors(
bond_list)
#debug_msg = "当前转债存量:{:.2f}, 低估转债存量:{:.2f}， 当前平均价格:{:.2f}, 当前溢价率{:.2f}".format(
# total_market, underrate_market, avg_price, avg_premium_ratio)
avg_prices.append(avg_price)
avg_premium_ratios.append(avg_premium_ratio)
total_markets.append(total_market)
underrate_markets.append(underrate_market)
days.append(day)
result = pd.DataFrame(
{
'total_markets': total_markets,
'underrate_markets': underrate_markets,
'avg_prices': avg_prices,
'avg_premium_ratios': avg_premium_ratios
},
index=days)
print(result)
return result
def initialize(context):
set_benchmark('000300.XSHG') # 设定比较基准指数
set_option('use_real_price', True) # 默认复权模式
# 设定融资/融券账户
set_subportfolios(
[SubPortfolioConfig(cash=context.portfolio.cash, type='stock_margin')])
# 设定交易成本
# set_order_cost(OrderCost(open_tax=0, close_tax=0.000, open_commission=0.0000, close_commission=0.0000, close_today_commission=0, min_commission=0), type='stock')
set_order_cost(OrderCost(open_tax=0,
close_tax=0.001,
open_commission=0.0003,
close_commission=0.0003,
close_today_commission=0,
min_commission=5),
type='stock')
# 设定融资/融券的利率和保证金比例
"""
set_option('margincash_interest_rate', 0.00)
set_option('margincash_margin_rate', 1.0)
set_option('marginsec_interest_rate', 0.00)
set_option('marginsec_margin_rate', 1.0)
"""
set_option('margincash_interest_rate', 0.00) # 事实上做多可以直接使用普通账户，所以不计这部分利率
set_option('margincash_margin_rate', 1.0)
set_option('marginsec_interest_rate', 0.10)
set_option('marginsec_margin_rate', 1.5)
# 设置定时运行函数
run_daily(before_market_open,
time='before_open',
reference_security='000300.XSHG')
run_daily(market_open, time='open', reference_security='000300.XSHG')
run_daily(after_market_close,
time='after_close',
reference_security='000300.XSHG')
# 记录交易日期及其索引
g.all_tdays = jqdata.get_all_trade_days()
g.date_xlat = dict([(doc[1], doc[0]) for doc in enumerate(g.all_tdays)])
# 设定若干全局变量
g.elapsed_days = 0
g.rebalance_today = False
g.new_long_pos = {}
g.new_short_pos = {}
def get_stock_beta_history_factors(self, interval=7):
"""
获取任意指数一段时间的历史 pe,pb 估值列表，通过计算当前的估值在历史估值的百分位，来判断当前市场的估值高低。
由于加权方式可能不同，可能公开的估值数据有差异，但用于判断估值相对高低没有问题
input：
interval: 计算指数估值的间隔天数，增加间隔时间可提高计算性能
output：
result: 指数历史估值的 DataFrame，index 为时间，列为pe，pb,roe
"""
all_days = get_all_trade_days()
pes = []
roes = []
pbs = []
days = []
begin = datetime.strptime(self._begin_date, '%Y-%m-%d').date()
end = datetime.strptime(self._end_date, '%Y-%m-%d').date()
i = 0
for day in all_days:
if (day <= begin or day >= end):
continue
i += 1
if (i % interval != 0):
continue
pe, pb, roe = self.get_stock_beta_factor(day)
if pe and pb and roe:
pes.append(pe)
pbs.append(pb)
roes.append(roe)
days.append(day)
result = pd.DataFrame({'pe': pes, 'pb': pbs, 'roe': roes}, index=days)
return result
set_option('margincash_margin_rate', 1.0)
set_option('marginsec_interest_rate', 0.00)
set_option('marginsec_margin_rate', 1.0)
"""
set_option('margincash_interest_rate', 0.00) # 事实上做多可以直接使用普通账户，所以不计这部分利率
set_option('margincash_margin_rate', 1.5)
set_option('marginsec_interest_rate', 0.10)
set_option('marginsec_margin_rate', 1.5)
# 设置定时运行函数
run_daily(before_market_open, time='before_open', reference_security='000300.XSHG')
run_daily(market_open, time='open', reference_security='000300.XSHG')
run_daily(after_market_close, time='after_close', reference_security='000300.XSHG')
# 记录交易日期及其索引
g.all_tdays = jqdata.get_all_trade_days()
g.date_xlat = dict([(doc[1], doc[0]) for doc in enumerate(g.all_tdays)])
# 设定若干全局变量
g.elapsed_days = 0
g.rebalance_today = False
g.new_long_pos = {}
g.new_short_pos = {}
# 在每个交易日开盘前运行
def before_market_open(context):
cur_dt = context.current_dt.date() # 当前日期
if g.elapsed_days % ADJUST_INTERVAL == 0: # 是否是调仓日
log.info("Rebalance at %s", cur_dt)
g.rebalance_today = True
idx = g.date_xlat[cur_dt]
def prepare_data(start_date, end_date, cut_date):
# 用沪深300做测试
test_stock = '000300.XSHG'
# 测试开始日期和终止日期
#start_date = datetime.date(2007, 1, 4)
#end_date = datetime.date(2017, 1, 3)
# 所有的交易日期
trading_days = jqdata.get_all_trade_days()
start_date_index = trading_days.tolist().index(start_date)
end_date_index = trading_days.tolist().index(end_date)
cut_date_index = trading_days.tolist().index(cut_date)
all_data = []
training_data = []
testing_data = []
feature_list = []
label_list = []
# 获取所有的数据
for index in range(start_date_index, end_date_index):
#计算特征值
start_day = trading_days[index - 35]
end_day = trading_days[index]
stock_data = get_price(test_stock,
start_date=start_day,
end_date=end_day,
frequency='daily',
fields=['close', 'high', 'low', 'volume'])
close_prices = stock_data['close'].values
high_prices = stock_data['high'].values
low_prices = stock_data['low'].values
volumes = stock_data['volume'].values
#通过数据计算指标
sma_data = talib.SMA(close_prices)[-1]
wma_data = talib.WMA(close_prices)[-1]
mom_data = talib.MOM(close_prices)[-1]
stck, stcd = talib.STOCH(high_prices, low_prices, close_prices)
stck_data = stck[-1]
stcd_data = stcd[-1]
macd, macdsignal, macdhist = talib.MACD(close_prices)
macd_data = macd[-1]
rsi_data = talib.RSI(close_prices, timeperiod=10)[-1]
willr_data = talib.WILLR(high_prices, low_prices, close_prices)[-1]
cci_data = talib.CCI(high_prices, low_prices, close_prices)[-1]
mfi_data = talib.MFI(high_prices, low_prices, close_prices,
volumes)[-1]
obv_data = talib.OBV(close_prices, volumes)[-1]
roc_data = talib.ROC(close_prices)[-1]
cmo_data = talib.CMO(close_prices)[-1]
# 训练数据或者测试数据的输入特征
features = []
features.append([sma_data]) #0
features.append([0])
features.append([wma_data]) #2
features.append([0])
features.append([mom_data]) #4
features.append([0])
features.append([stck_data]) #6
features.append([0])
features.append([stcd_data]) #8
features.append([0])
features.append([macd_data]) #10
features.append([0])
features.append([rsi_data]) #12
features.append([0])
features.append([willr_data]) #14
features.append([0])
features.append([cci_data]) #16
features.append([0])
features.append([mfi_data]) #18
features.append([0])
features.append([obv_data]) #20
features.append([0])
features.append([roc_data]) #22
features.append([0])
features.append([cmo_data]) #24
features.append([0])
features.append([close_prices[-1]])
features = np.array(features)
# 训练数据或者测试的标签数据，就是涨或者跌，涨用[1,0]，平或者跌用[0,1]
label = [[0], [1]]
if close_prices[-1] > close_prices[-2]:
label = [[1], [0]]
elif close_prices[-1] < close_prices[-2]:
label = [[0], [1]]
label = np.array(label)
feature_list.append(features)
label_list.append(label)
# 连续数值离散化
# 从后面开始向前面遍历，第一行数据需要舍弃,range只包含第一个元素，不包含第二个元素
for index in range(len(feature_list) - 1, 0, -1):
# SMA
if feature_list[index][0][0] < feature_list[index][-1][0]:
feature_list[index][0][0] = 1
feature_list[index][1][0] = 0
else:
feature_list[index][0][0] = 0
feature_list[index][1][0] = 1
# WMA
if feature_list[index][2][0] < feature_list[index][-1][0]:
feature_list[index][2][0] = 1
feature_list[index][3][0] = 0
else:
feature_list[index][2][0] = 0
feature_list[index][3][0] = 1
# MOM
if feature_list[index][4][0] > 0:
feature_list[index][4][0] = 1
feature_list[index][5][0] = 0
else:
feature_list[index][4][0] = 0
feature_list[index][5][0] = 1
# STCK
if feature_list[index][6][0] > feature_list[index - 1][6][0]:
feature_list[index][6][0] = 1
feature_list[index][7][0] = 0
else:
feature_list[index][6][0] = 0
feature_list[index][7][0] = 1
# STCD
if feature_list[index][8][0] > feature_list[index - 1][8][0]:
feature_list[index][8][0] = 1
feature_list[index][9][0] = 0
else:
feature_list[index][8][0] = 0
feature_list[index][9][0] = 1
# MACD
if feature_list[index][10][0] > feature_list[index - 1][10][0]:
feature_list[index][10][0] = 1
feature_list[index][11][0] = 0
else:
feature_list[index][10][0] = 0
feature_list[index][11][0] = 1
# RSI
if feature_list[index][12][0] > 70:
feature_list[index][12][0] = 0
feature_list[index][13][0] = 1
elif feature_list[index][12][0] < 30:
feature_list[index][12][0] = 1
feature_list[index][13][0] = 0
else:
if feature_list[index][12][0] > feature_list[index - 1][12][0]:
feature_list[index][12][0] = 1
feature_list[index][13][0] = 0
else:
feature_list[index][12][0] = 0
feature_list[index][13][0] = 1
# WILLR
if feature_list[index][14][0] > feature_list[index - 1][14][0]:
feature_list[index][14][0] = 1
feature_list[index][15][0] = 0
else:
feature_list[index][14][0] = 0
feature_list[index][15][0] = 1
# CCI
if feature_list[index][16][0] > 200:
feature_list[index][16][0] = 0
feature_list[index][17][0] = 1
elif feature_list[index][16][0] < -200:
feature_list[index][16][0] = 1
feature_list[index][17][0] = 0
else:
if feature_list[index][16][0] > feature_list[index - 1][16][0]:
feature_list[index][16][0] = 1
feature_list[index][17][0] = 0
else:
feature_list[index][16][0] = 0
feature_list[index][17][0] = 1
# MFI
if feature_list[index][18][0] > 90:
feature_list[index][18][0] = 0
feature_list[index][19][0] = 1
elif feature_list[index][18][0] < 10:
feature_list[index][18][0] = 1
feature_list[index][19][0] = 0
else:
if feature_list[index][18][0] > feature_list[index - 1][18][0]:
feature_list[index][18][0] = 1
feature_list[index][19][0] = 0
else:
feature_list[index][18][0] = 0
feature_list[index][19][0] = 1
# OBV
if feature_list[index][20][0] > feature_list[index - 1][20][0]:
feature_list[index][20][0] = 1
feature_list[index][21][0] = 0
else:
feature_list[index][20][0] = 0
feature_list[index][21][0] = 1
# ROC
if feature_list[index][22][0] > 0:
feature_list[index][22][0] = 1
feature_list[index][23][0] = 0
else:
feature_list[index][22][0] = 0
feature_list[index][23][0] = 1
# CMO
if feature_list[index][24][0] > 50:
feature_list[index][24][0] = 0
feature_list[index][25][0] = 1
elif feature_list[index][24][0] < -50:
feature_list[index][24][0] = 1
feature_list[index][25][0] = 0
else:
if feature_list[index][24][0] > feature_list[index - 1][24][0]:
feature_list[index][24][0] = 1
feature_list[index][25][0] = 0
else:
feature_list[index][24][0] = 0
feature_list[index][25][0] = 1
# 删除价格
feature_list[index] = np.delete(feature_list[index], -1, axis=0)
for i in range(1, len(feature_list)):
sample_term = (feature_list[i], label_list[i])
if 1 + i + start_date_index < cut_date_index:
training_data.append(sample_term)
elif 1 + i + start_date_index >= cut_date_index:
testing_data.append(sample_term)
print("Data prepared. training data size: {0}, testing data size: {1}".
format(len(training_data), len(testing_data)))
return training_data, testing_data
