w********e
发帖数: 944 | 1 看到有些job descriptions都要求有decision tree的知识.大家可不可以说说这方面的
知识
需要多深? |
C*********g
发帖数: 3728 | 2 easy stuff
and it'll be out of the market soon.
【在 w********e 的大作中提到】
: 看到有些job descriptions都要求有decision tree的知识.大家可不可以说说这方面的
: 知识
: 需要多深?
|
w********e
发帖数: 944 | 3 愿闻其详.
【在 C*********g 的大作中提到】
: easy stuff
: and it'll be out of the market soon.
|
g********r
发帖数: 8017 | 4 really？ Are you considering random forest and boosted trees?
【在 C*********g 的大作中提到】
: easy stuff
: and it'll be out of the market soon.
|
w********e
发帖数: 944 | 5 This is what extactly puzzled me. Does 'decision tree' in the job
description refer to simple decision tree or a whole bunch of tree stuff
like random forest bla bla...?
BTW, the job here refers to general positions like those in risk management,
not those tough research stuffs. |
s***x
发帖数: 293 | 6 Too naive to tell the truth! Even the simplest decision tree is sometimes very useful
such as CART.
Sorry for my 'maofan', just for your point instead of you.
【在 C*********g 的大作中提到】
: easy stuff
: and it'll be out of the market soon.
|
g********r
发帖数: 8017 | 7 Gee. Risk management uses a lot of data mining. I don't think any serious
risk management job will use the
naive tree. The difference of performance between CART and forests could be
measured by miles in many
situations. Doesn't hurt to know more. The intuition behind random forest
and tree boosting is straight
forward.
management,
【在 w********e 的大作中提到】
: This is what extactly puzzled me. Does 'decision tree' in the job
: description refer to simple decision tree or a whole bunch of tree stuff
: like random forest bla bla...?
: BTW, the job here refers to general positions like those in risk management,
: not those tough research stuffs.
|
w********e
发帖数: 944 | 8 You sounds an expert in this field. Can you elaborate on this topic a little
more?
Thanks a lot.
be
【在 g********r 的大作中提到】
: Gee. Risk management uses a lot of data mining. I don't think any serious
: risk management job will use the
: naive tree. The difference of performance between CART and forests could be
: measured by miles in many
: situations. Doesn't hurt to know more. The intuition behind random forest
: and tree boosting is straight
: forward.
:
: management,
|
|
g********r
发帖数: 8017 | 9 I'm not. Just look at Hastie's book. The new edition came out this year.
Quite some tree stuff. The authors are
very good in doing intuitive explanations. You don't need to study a lot of
math to sound like you know it in a
phone interview.
little
【在 w********e 的大作中提到】
: You sounds an expert in this field. Can you elaborate on this topic a little
: more?
: Thanks a lot.
:
: be
|
w********e
发帖数: 944 | 10 I do realize that models can be very complicated in research. Just don't
have sense about how it is applied in industry.
of
【在 g********r 的大作中提到】
: I'm not. Just look at Hastie's book. The new edition came out this year.
: Quite some tree stuff. The authors are
: very good in doing intuitive explanations. You don't need to study a lot of
: math to sound like you know it in a
: phone interview.
:
: little
|
g********r
发帖数: 8017 | 11 They are complicated for a reason. If a model has no good sensitivity/
specificity, or it is not robust, it won't
do well in industry either. Random forests can not only give relatively
accurate predictions, but it can be used
to assess the relative importance of predictors, which I suppose is
important in risk models. A simple tree
cannot achieve those.
It is not true that industry use simpler models, especially when it comes to banks and mortgage companies.
【在 w********e 的大作中提到】
: I do realize that models can be very complicated in research. Just don't
: have sense about how it is applied in industry.
:
: of
|
