原始因子處理之手寫標準化函數

一、標準化

代碼實現（z_score方法標準化）

def standardlize(data, inf2nan=True, axis=1):
    '''
    參數
    -----------
    data: pd.Series/pd.DataFrame/np.array, 待標準化的序列
    inf2nan: 是否將 np.inf 和 -np.inf 替換成 np.nan。默認爲 True
    axis=1: 在 data 爲 pd.DataFrame 時使用，如果 series 爲 pd.DataFrame，沿哪個方向做標準化。0 爲對每列做標準化，1 爲對每行做標準化
    返回
    -----------
    標準化後的因子數據
    '''
    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值有誤')
            
    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類型')
        return