def __init__(
self,
factor: Union[pd.Series,
pd.DataFrame],
prices: Union[pd.Series,
pd.DataFrame,
Callable],
groupby: Union[pd.Series,
pd.DataFrame,
Callable],
stock_start_date: Union[pd.Series,
pd.DataFrame,
Callable],
weights: Union[float,
pd.Series,
pd.DataFrame,
Callable] = 1.0,
frequence: str = "DAY",
quantiles: Union[int,
Tuple[float],
List[float]] = 5,
bins: Union[int,
Tuple[float],
List[float]] = None,
periods: Union[int,
Tuple[int],
List[int]] = (1,
5,
10),
binning_by_group: bool = False,
max_loss: float = 0.25,
zero_aware: bool = False,
):
"""
初始化输入
说明：
- quantiles 与 bins 只能有一个为非空
"""
self.factor = preprocess.QA_fmt_factor(factor)
self.prices = prices
self.groupby = groupby
self.stock_start_date = stock_start_date
self.weights = weights
self.frequence = utils.get_frequence(frequence)
self.quantiles = quantiles
self.bins = bins
if isinstance(periods, int):
periods = (periods,)
self.periods = periods
self.binning_by_group = binning_by_group
self.max_loss = max_loss
self.zero_aware = zero_aware
# 因子加工与添加因子远期收益
self.__gen_clean_factor_and_forward_returns()
def __init__(
self,
jq_username: str = None,
jq_password: str = None,
price_type: str = "close", # 价格类型
fq: str = "qfq", # 复权方式
factor_time_range: list = None,
industry_cls: str = "sw_l1", # 行业分类
industry_data: Union[dict, pd.Series, pd.DataFrame] = None, # 行业信息
weight_cls: str = "avg", # 权重分类
weight_data: Union[dict, pd.Series, pd.DataFrame] = None, # 权重信息
frequence: str = "DAY", # 频率
detailed: bool = False, # 是否按日期进行行业查询
):
"""
根据输入的参数进行初始化
说明
---
如果需要使用自定义的行业，权重信息，需要设置 industry_cls, weight_cls 为 None, 然后再输入 industry_data, weight_data
参数
---
:param jq_username: 聚宽账号
:param jq_password: 聚宽密码
:param price_type: 计算因子收益时，使用价格数据类型，支持 ['open', 'high', 'low', 'close', 'avg'], 默认为 'close'
:param fq: 价格数据的复权方式
:param factor_time_range: 因子对应的时间范围
:param industry_cls: 行业类型，目前支持
- "sw_l1": 申万一级
- "sw_l2": 申万二级
- "sw_l3": 申万三级
- "jq_l1": 聚宽一级
- "jq_l2": 聚宽二级
- "zjw": 证监会
:param indsutry_data: 行业数据 [可选], 为 industry_cls 只能有一个为非 None
:param weight_cls: 权重类型，目前支持
- "avg": 等权重
- "mktcap": 总市值加权
- "cmktcap": 流通市值加权
- "ln_mktcap": 对数市值加权
- "ln_cmktcap": 对数流通市值加权
- "sqrt_mktcap": 市值平方根加权
- "sqrt_cmktcap": 流通市值平方根加权
:param weight_data: 权重数据 [可选], 与 weight_cls 两种只能有一个为非 None
:param frequence: 频率
:param detailed: 是否使用详细的按日期分类行业信息，默认使用 end_date 的行业数据
"""
jqdatasdk.auth(jq_username, jq_password)
if price_type is None:
price_type = "close"
warnings.warn("没有指定使用的价格类型，采用收盘价进行计算", UserWarning)
assert price_type.lower() in PRICE_TYPE
self.price_type = price_type
if fq is not None:
assert fq.lower() in FQ_TYPE
else:
warnings.warn("使用不复权的价格数据进行计算，结果可能受到分红配股等影响而不准确")
self.fq = fq
self.start_time = min(factor_time_range)
self.end_time = max(factor_time_range)
if industry_cls is not None:
assert industry_cls.lower() in INDUSTRY_CLS
# 如果输入了指定行业分类方式，则忽略行业数据数据
self.industry_cls = industry_cls
self.industry_data = None
else:
self.industry_cls = industry_cls
self.industry_data = industry_data
if (self.industry_cls is None) and (self.industry_data is None):
warnings.warn("没有指定行业分类方式，也没有输入行业数据", UserWarning)
if weight_cls is not None:
assert weight_cls.lower() in WEIGHT_CLS
self.weight_cls = weight_cls
self.weight_data = None
else:
self.weight_cls = weight_cls
self.weight_data = weight_data
if (self.weight_cls is None) and (self.weight_data is None):
warnings.warn("没有指定加权方式，也没有输入加权数据，默认采用等权重方式", UserWarning)
self.weight_cls = "avg"
self.weight_data = None
self.frequence = utils.get_frequence(frequence)
self.detailed = detailed
def get_prices(
self,
code_list: Union[str, Tuple[str], List[str]] = None,
start_time: Union[str, datetime.datetime] = None,
end_time: Union[str, datetime.datetime] = None,
fq: str = None,
frequence: str = None,
price_type: str = None,
):
"""
价格数据获取接口，单因子输入后，可以通过单因子获取股票代码，时间等参数信息
参数
---
:param code_list: 股票代码
:param start_time: 起始时间
:param end_time: 截止时间
:param fq: 复权方式
:param frequence: 时间频率
"""
# 1. 股票池
if isinstance(code_list, tuple):
code_list = list(code_list)
# 2. 时间频率
if not frequence:
frequence = self.frequence
frequence = utils.get_frequence(frequence)
if not start_time:
start_time = self.start_time
if not end_time:
end_time = self.end_time
if (not start_time) or (not end_time):
raise ValueError("价格获取接口需要指定起始时间与结束时间")
start_time = str(pd.Timestamp(start_time))[:19]
end_time = str(pd.Timestamp(end_time))[:19]
data = QA_fetch_stock_day_adv(code=code_list,
start=start_time,
end=end_time)
index_data = QA_fetch_index_day_adv(code="000001",
start=start_time,
end=end_time)
# 3. 复权
if not fq:
fq = self.fq
if not fq:
data = data
elif fq.lower() in ["pre", "qfq", "前复权"]:
data = data.to_qfq()
elif fq.lower() in ["post", "hfq", "后复权"]:
data = data.to_hfq()
elif fq.lower() in ["none", "bfq", "不复权"]:
data = data
# 4. 重采样
# 考虑到停牌退市等原因，重采样会有异常值，即日期与我们需要的日期不一致
# 这里采用指数作为基准，对重采样数据进行再处理
# 对于停牌数据缺失，采用前值作为填充
if frequence == "1d":
data = data.data.unstack().ffill().stack()
else:
index_data = index_data.resample(frequence).unstack()
data = data.resample(frequence).unstack().ffill()
data = data.reindex(index_data.index).stack()
if frequence == '1q':
data.index = data.index.map(lambda x:
(utils.QA_fmt_quarter(x[0]), x[1]))
# 5. 价格类型
if not price_type:
price_type = self.price_type
if price_type.lower() is "avg":
avg = data["amount"] / data["volume"] / 100.0
return avg.unstack()
return data[price_type.lower()].unstack()
def __init__(
self,
factor: Union[pd.Series,
pd.DataFrame],
prices: Union[pd.Series,
pd.DataFrame,
Callable],
groupby: Union[pd.Series,
pd.DataFrame,
Callable],
stock_start_date: Union[pd.Series,
pd.DataFrame,
Callable],
weights: Union[float,
pd.Series,
pd.DataFrame,
Callable] = 1.0,
frequence: str = "DAY",
quantiles: Union[int,
Tuple[float],
List[float]] = 5,
bins: Union[int,
Tuple[float],
List[float]] = None,
periods: Union[int,
Tuple[int],
List[int]] = (1,
5,
10),
binning_by_group: bool = False,
max_loss: float = 0.25,
zero_aware: bool = False,
):
"""
初始化输入
参数
---
:param factor: 即经过我们处理过的单因子数据
:param prices: 价格数据
:param groupby: 行业数据
:param stock_start_date: 上市时间
:param weights: 因子加权数据
:param frequence: 因子频率，如果是季线，相应的为 `1q`, 日线，则为 `1d` 等
:param quantiles: 分位处理，既可以是 int 值，将因子等分，也可以是 0 到 1 之间的非等分区间
:param bins: 分位处理，与 `quantiles` 只能有一个为非空
:param periods: 用于计算因子远期收益的期数
:param binning_by_group: 是否按照行业分别进行分位处理
:param max_loss: 在单因子处理时，能够忍受的最大的因子损失 (丢弃的 NaN 值比例)
:param zero_aware: 是否按照因子正负值分别进行分位处理
说明：
- quantiles 与 bins 只能有一个为非空
"""
self.factor = preprocess.QA_fmt_factor(factor)
self.prices = prices
self.groupby = groupby
self.stock_start_date = stock_start_date
self.weights = weights
self.frequence = utils.get_frequence(frequence)
self.quantiles = quantiles
self.bins = bins
if isinstance(periods, int):
periods = (periods,)
self.periods = periods
self.binning_by_group = binning_by_group
self.max_loss = max_loss
self.zero_aware = zero_aware
# 因子加工与添加因子远期收益
self.__gen_clean_factor_and_forward_returns()
