作者daze (一期一会)
看板CFP
标题Re: [问卷] 智能理财使用意愿调查
时间Fri Mar 12 09:19:44 2021
1F:推 makio: 想搭版问,金融商品和顾客风险属性的组合有限,为何要用到A 03/12 01:32
2F:→ makio: I呢?感觉传统的判断式就能达成了 03/12 01:32
对於数学可以给出显式解的问题
没有使用AI的必要
很多隐式解也可以基於数值方法求近似解
再者,基於Deep learning的AI需要feedback
AI理财的feedback是什麽?
如果是三十年後顾客的投资成果
这个AI怎麽可能能够成长得起来?
要嘛这些「AI理财」只是marketing手法
实际上根本没有deep learning
要嘛所用的feedback根本不是基於客户的利益
举例来说
如果把deep learning的feedback
放在客户会不会接受被推荐的标的
或者放在怎麽样推荐能让客户心甘情愿交出更多手续费
这样就有即时feedback可以用来喂给AI了
如果把feedback放在「客户1个月後的绩效」
这恐怕也与客户真正的利益无关
反而可能导致不良的投资行为
======
求出客户的风险属性
搞不好反而是AI可以着力的点
像这里 (
https://tinyurl.com/wz67549 )估计的相对风险趋避系数
是基於constant relative risk aversion假设
且人难自知
AI透过各种问题与回馈的收集
或许能对效用函数进行分段估计
能更准确的反映个人的风险属性也说不定
--
You got to know when to hold 'em, know when to fold 'em, Know when to walk away and know when to run.
You never count your money when you're sittin' at the table. There'll be time enough for countin' when the dealin's done.
'Cause ev'ry hand's a winner and ev'ry hand's a loser, And the best that you can hope for is to die in your sleep."
now Ev'ry gambler knows that the secret to survivin' Is knowin' what to throw away and knowing what to keep.
--
※ 发信站: 批踢踢实业坊(ptt.cc), 来自: 114.39.53.4 (台湾)
※ 文章网址: https://webptt.com/cn.aspx?n=bbs/CFP/M.1615511990.A.F95.html
※ 编辑: daze (60.249.243.225 台湾), 03/12/2021 09:55:05
3F:推 makio: 而且如果把历史资料拿去深度学习,会变成悖论吧...?所以 03/12 19:58
4F:→ makio: 应该是做不到预测的。我的理解不知道对不对。只能依照风险 03/12 19:58
5F:→ makio: 属性做投资组合的配对而已吧。 03/12 19:58