原始因子处理之手写标准化函数

一、标准化

代码实现（z_score方法标准化）

def standardlize(data, inf2nan=True, axis=1):
    '''
    参数
    -----------
    data: pd.Series/pd.DataFrame/np.array, 待标准化的序列
    inf2nan: 是否将 np.inf 和 -np.inf 替换成 np.nan。默认为 True
    axis=1: 在 data 为 pd.DataFrame 时使用，如果 series 为 pd.DataFrame，沿哪个方向做标准化。0 为对每列做标准化，1 为对每行做标准化
    返回
    -----------
    标准化后的因子数据
    '''
    if isinstance(data,pd.DataFrame):
        value = data.copy()
        if axis==1:
            long = value.shape[0]
            for i in range(long):
                s = value.iloc[i,:]
                if inf2nan==True:
                    s[np.isinf(s)]=np.nan
                    mean = np.mean(s.dropna())
                    std = np.std(s.dropna(),ddof=1)
                    value.iloc[i,:] = (s-mean)/std            
                else: 
                    s1 = s[~np.isinf(s)]
                    mean = np.mean(s1)
                    std = np.std(s1,ddof=1)
                    value.iloc[i,:] = (s-mean)/std
            return value
        elif axis==0:
            width = value.shape[1]
            for j in range(width):
                s = value.iloc[:,j]
                if inf2nan==True:
                    s[np.isinf(s)]=np.nan
                    mean = np.mean(s.dropna())
                    std = np.std(s.dropna(),ddof=1)
                    value.iloc[:,j] = (s-mean)/std            
                else: 
                    s1 = s[~np.isinf(s)]
                    mean = np.mean(s1)
                    std = np.std(s1,ddof=1)
                    value.iloc[:,j] = (s-mean)/std
            return value
        else:
            return('axis值有误')
            
    elif isinstance(data,pd.Series):
        value = data.copy()
        if inf2nan==True:
            value[np.isinf(value)]=np.nan
            mean = np.mean(value.dropna())
            std = np.std(value.dropna(),ddof=1)
            value = (value-mean)/std
            return value
        else: 
            s = value[~np.isinf(value)]
            mean = np.mean(s)
            std = np.std(s,ddof=1)
            value = (value-mean)/std
            return value
    else:
        print('data不是pd.Series和pd.DataFrame类型')
        return