def set_factors_labels(self, price=None, ta_list=None, lags=None):
"""
根据价格数据及选定的指标计算特征
:return: the factors array DataFrame
"""
ta_fun_names = ta.get_functions()
df_price = price if price else self.price_data
ta_list = ta_list if ta_list else self.dic_factors
lags = lags if lags else self.ta_lags
ta_factors = df_price.copy()
ii = 1
for j in ta_list:
if j[0] in ta_fun_names:
ta_fun = Function(j[0])
if j[1] == {}:
ta_fun = ta_fun
else:
dic_param = dict(ta_fun.parameters)
for key1 in dic_param:
dic_param[key1] = j[1][key1]
ta_fun.parameters = dic_param
ta_value = ta_fun(df_price)
if ta_value.size > len(ta_value):
ta_factors["%s_%d" % (j[0], ii)] = ta_value.ix[:, 0]
else:
ta_factors["%s_%d" % (j[0], ii)] = ta_value
ii += 1
ta_factors['ACC'] = TSSB_TA(df_price).ACC()
ta_factors['ADOSC'] = TSSB_TA(df_price).ADOSC()
# 计算lags指标
if lags > 0:
ret = pd.DataFrame(index=ta_factors.index)
columns = ta_factors.columns
for col in columns:
for lag in xrange(0, lags):
ret[col + "_%s" % str(lag+1)] = ta_factors[col].shift(lag)
else:
ret = ta_factors
# 定义标签: 上涨为1， 下跌为-1
if self.predict_type == 1:
labels = df_price['close'].pct_change().shift(-1) # 预测明收盘与今收盘的涨跌
else:
labels = (df_price['close'] - df_price['open']).shift(-1) # 预测明收盘与明开盘的涨跌
labels[labels >= 0] = 1
labels[labels < 0] = 0
ret = ret.dropna()
labels = labels[ret.index]
self.ta_factors, self.labels = ret, labels
return ret, labels
def abstract_example():
sma = Function("sma")
input_arrays = sma.get_input_arrays()
for key in input_arrays.keys():
input_arrays[key] = idata
sma.set_input_arrays(input_arrays)
odata = sma(30) # timePeriod=30, specified as an arg
bbands = Function("bbands", input_arrays)
bbands.parameters = {"timeperiod": 20, "nbdevup": 2, "nbdevdn": 2}
upper, middle, lower = bbands() # multiple output values unpacked (these will always have the correct order)
kama = Function("kama").run(input_arrays) # alternative run() calling method.
plot(odata, upper, middle, lower, kama)
def abstract_example():
sma = Function('sma')
input_arrays = sma.get_input_arrays()
for key in input_arrays.keys():
input_arrays[key] = idata
sma.set_input_arrays(input_arrays)
odata = sma(30) # timePeriod=30, specified as an arg
bbands = Function('bbands', input_arrays)
bbands.parameters = {'timeperiod': 20, 'nbdevup': 2, 'nbdevdn': 2}
upper, middle, lower = bbands(
) # multiple output values unpacked (these will always have the correct order)
kama = Function('kama').run(
input_arrays) # alternative run() calling method.
plot(odata, upper, middle, lower, kama)
def __ta_look_back(self):
"""
获取所需回溯历史数据的长度，计算技术指标时需要n天前的数据，同时预测涨跌时又要windows_size的历史指标
:return:
"""
ta_fun_names = ta.get_functions()
lk_list = []
for j in self.dic_factors:
if j[0] in ta_fun_names:
ta_fun = Function(j[0])
if j[1] == {}:
continue
else:
dic_param = dict(ta_fun.parameters)
for key1 in dic_param:
dic_param[key1] = j[1][key1]
ta_fun.parameters = dic_param
lk = ta_fun.lookback
lk_list.append(lk)
else:
pass
ret = max(lk_list)
return ret
