分享两个刚刚面试的modeling问题 - 未名空间MITBBS历史存档

分享两个刚刚面试的modeling问题# DataSciences - 数据科学

b*g2018-10-23 07:10

1 楼

Display advertising is a billion dollar level effort
and one of the central uses of machine learning on the Internet.
However, its data and methods are usually kept under lock and key.
Amex has a worth of data collected from different channels throughout
customer’s entire web browsing journey.
1) Given a user and the product page he is visiting,
and given that there are five products displayed on this page,
predict which he/she will click on
2) Please also predict a user’s click journey.
Specifically, given a sequence of k clicks, predict the next,
or k+1 click

f*02018-10-23 07:10

2 楼

1). classification problem. Multinomial logistic regression?
2). markov chain problem. Transition matrix model? or Survival Analysis?

m*r2018-10-23 07:10

3 楼

第二个比较难。第一个也得看输入都是些什么变量。如果题目真是amex卖产品， 5张
信用卡让用户选，不觉得和在线广告有什么关系。
在线广告是个bidding process, 你弄个高级算法，算出来比别人晚了两毫秒，就被别
人抢了先了。

E*e2018-10-23 07:10

4 楼

Association rule or one of recommendation algorithms?

【在 b**********g 的大作中提到】

: Display advertising is a billion dollar level effort
: and one of the central uses of machine learning on the Internet.
: However, its data and methods are usually kept under lock and key.
: Amex has a worth of data collected from different channels throughout
: customer’s entire web browsing journey.
: 1) Given a user and the product page he is visiting,
: and given that there are five products displayed on this page,
: predict which he/she will click on
: 2) Please also predict a user’s click journey.
: Specifically, given a sequence of k clicks, predict the next,

E*e2018-10-23 07:10

5 楼

Association rule or one of recommendation algorithms?

【在 b**********g 的大作中提到】

: Display advertising is a billion dollar level effort
: and one of the central uses of machine learning on the Internet.
: However, its data and methods are usually kept under lock and key.
: Amex has a worth of data collected from different channels throughout
: customer’s entire web browsing journey.
: 1) Given a user and the product page he is visiting,
: and given that there are five products displayed on this page,
: predict which he/she will click on
: 2) Please also predict a user’s click journey.
: Specifically, given a sequence of k clicks, predict the next,

b*g2018-10-23 07:10

6 楼

Display advertising is a billion dollar level effort
and one of the central uses of machine learning on the Internet.
However, its data and methods are usually kept under lock and key.
Amex has a worth of data collected from different channels throughout
customer’s entire web browsing journey.
1) Given a user and the product page he is visiting,
and given that there are five products displayed on this page,
predict which he/she will click on
2) Please also predict a user’s click journey.
Specifically, given a sequence of k clicks, predict the next,
or k+1 click

f*02018-10-23 07:10

7 楼

1). classification problem. Multinomial logistic regression?
2). markov chain problem. Transition matrix model? or Survival Analysis?

m*r2018-10-23 07:10

8 楼

第二个比较难。第一个也得看输入都是些什么变量。如果题目真是amex卖产品， 5张
信用卡让用户选，不觉得和在线广告有什么关系。
在线广告是个bidding process, 你弄个高级算法，算出来比别人晚了两毫秒，就被别
人抢了先了。

E*e2018-10-23 07:10

9 楼

Association rule or one of recommendation algorithms?

【在 b**********g 的大作中提到】

: Display advertising is a billion dollar level effort
: and one of the central uses of machine learning on the Internet.
: However, its data and methods are usually kept under lock and key.
: Amex has a worth of data collected from different channels throughout
: customer’s entire web browsing journey.
: 1) Given a user and the product page he is visiting,
: and given that there are five products displayed on this page,
: predict which he/she will click on
: 2) Please also predict a user’s click journey.
: Specifically, given a sequence of k clicks, predict the next,

E*e2018-10-23 07:10

10 楼

Association rule or one of recommendation algorithms?

【在 b**********g 的大作中提到】

: Display advertising is a billion dollar level effort
: and one of the central uses of machine learning on the Internet.
: However, its data and methods are usually kept under lock and key.
: Amex has a worth of data collected from different channels throughout
: customer’s entire web browsing journey.
: 1) Given a user and the product page he is visiting,
: and given that there are five products displayed on this page,
: predict which he/she will click on
: 2) Please also predict a user’s click journey.
: Specifically, given a sequence of k clicks, predict the next,

s*h2018-10-23 07:10

11 楼

Thanks for sharing.

a*o2018-10-23 07:10

12 楼

Obviouly the 1st question is emplying Classfication and the 2nd is applying
Prediction function.
While I don't think the second one is more difficult than that first one.
What's more, NN is much better in terms of the prediction precise and time-
consuming than Markov chain.
There are more advanced algorithms which could be employed here. Please do
research.

b*g2018-10-23 07:10

13 楼

当时秒挂了，谢谢大家的讨论