PayPal 数据科学家电面+昂赛

电面

Resume + a fraud model case

What’s the advantage of random forest over GBM?

• Less hyper parameters to tune
  
  

• RF shows very low variance
• Less likely to overfit. The more trees in RF, the better. That’s because the multitude of trees serves to reduce variance. Each tree fits, or overfits, a part of the training set, and in the end their errors cancel out, at least partially. Random forests do overfit, just compare the error on train and validation sets.
• easier to parallelize
  
  

Fraud case: given transactions 2010 ~ 2017 and 10k fields, build a model to detect fraud

• Explore data, plot one way correlation
• Feature selection information gain
  您好！
  本帖隐藏的内容需要积分高于 188 才可浏览
  您当前积分为 0。
  使用VIP即刻解锁阅读权限或查看其他获取积分的方式

  游客，您好！
  本帖隐藏的内容需要积分高于 188 才可浏览
  您当前积分为 0。
  VIP即刻解锁阅读权限 或 查看其他获取积分的方式
  Unlock interview details and practice with AI
  Curated Interview Questions from Top Companies
  
  taframe, numpy ndarray, or collections
  

Python & SQL

• Find median in 2 sorted arrays in O(logN)
• SQL or python: paypal table - (trxn_id, sender_id, receiver_id, trxn_amt). How will you present user engagement? Count number of trxns per sender. What if there are users with no trxn, need an additional table of all user_id
  
  

请问楼主，是找内推的么？还是recruiter找上门的？

daguanyuan 发表于 2018-1-25 07:41
请问楼主，是找内推的么？还是recruiter找上门的？

是内推的

chongchong90 发表于 2018-1-25 09:38
是内推的

可以问问楼主面的是什么position吗？看起来考了很多Nlp的东西？

How will you present user engagement? 请我楼主这个问题应该如何考虑呢？ the number of trxn and trxn_amount 是不是都要考虑的

楼主请问user engagement为什么是Count number of trxns per sender而不是number of trxns per user呀？

数据科学家面得好难啊

这是故意难为你吧？median of two sorted array…