年化422%，回撤7%，夏普比5.4| Deap因子挖掘新增qlib因子库，附python代码

原创内容第1011篇，专注AGI+，AI量化投资、个人成长与财富自由。

今日推荐策略：年化422%，回撤7%，夏普比5.4。

因子挖掘：

新增qlib的因子库：

复制代码

import numpy as npimport pandas as pd
import numpy as npimport pandas as pd

def sign(se: pd.Series):    return np.sign(se)

def signed_power(se: pd.Series, a):    return np.where(se < 0, -np.abs(se) ** a, np.abs(se) ** a)
def log(se: pd.Series):    return np.log(se)

def abs(se:pd.Series):    return np.abs(se)

def scale(x:pd.Series, a:int=1):    """    Scales the array x such that the sum of the absolute values equals a.
    Parameters:    x (array-like): The input array to be scaled.    a (float, optional): The target sum of absolute values. Default is 1.
    Returns:    numpy.ndarray: The scaled array.    """    import numpy as np    x = np.array(x)  # 确保输入是numpy数组    sum_abs_x = np.sum(np.abs(x))  # 计算x的绝对值之和    if sum_abs_x == 0:        raise ValueError("The sum of absolute values of x is zero, cannot scale by a non-zero value.")    scale_factor = a / sum_abs_x  # 计算缩放因子    return x * scale_factor  # 应用缩放因子
def decay_linear(series:pd.Series, window:int):    """    对输入的时间序列进行线性衰减。
    :param series: 输入的时间序列。    :param window: 衰减的窗口大小。    :return: 衰减后的序列。    """    weights = np.arange(1, window + 1)    decay = np.convolve(series, weights, 'valid') / np.sum(weights)    return decay

def delay(se: pd.Series, periods:int=5):    return se.shift(periods=periods)


def delta(se: pd.Series, periods:int=20):    se_result = se - se.shift(periods=periods)    return se_result

def ts_min(se: pd.Series, periods:int=5):    return se.rolling(window=periods).min()

def ts_max(se: pd.Series, periods:int=5):    return se.rolling(window=periods).max()


def ts_argmin(se: pd.Series, periods:int=5):    return se.rolling(periods, min_periods=2).apply(lambda x: x.argmin())

def ts_argmax(se: pd.Series, periods:int=5):    return se.rolling(periods, min_periods=2).apply(lambda x: x.argmax())

def stddev(se:pd.Series, periods:int=5):    return se.rolling(window=periods).std()

def ts_rank(se: pd.Series, periods:int=9):    ret = se.rolling(window=periods).rank(pct=True)    return ret

def sum(se: pd.Series, N:int):    ret = se.rolling(N).sum()    ret.name = 'sum_{}'.format(N)    return ret

def shift(se: pd.Series, N:int):    return se.shift(N)

def roc(se: pd.Series, N:int):    return se / shift(se, N) - 1

def product(se: pd.Series, d:int):    return se.rolling(window=d).apply(np.product)

def zscore(se: pd.Series, N:int):    def _zscore(x):
        try:            x.dropna(inplace=True)            #print('sub', x)            value = (x[-1] - x.mean()) / x.std()            if value:                return value        except:            return -1
    ret = se.rolling(window=N).apply(lambda x: _zscore(x))    return ret