中级农民
- 积分
- 107
- 大米
- 颗
- 鳄梨
- 个
- 水井
- 尺
- 蓝莓
- 颗
- 萝卜
- 根
- 小米
- 粒
- 学分
- 个
- 注册时间
- 2017-11-1
- 最后登录
- 1970-1-1
|
Looking for BA/DA/DS roles
记录我每天干了什么并进行小总结 方便自己之后浏览~
技术:
Python
R
SQL: take-home challenge: Q1: 没明白题目意思--- 用了lag /lead & row_number() 这样的window function, 注意用完之后到底是consecutive actions 到底是在什么rank下
Q2 较简单,但是我用了笨方法,需要改进 (case when)
Statistics
ML: Imbalanced data:
1. What is imbalanced data:
The dataset has multi-categories, which are not equally. For example, there are only limited numbers of fraud transactions in daily transactions. The fraud and non-fraud data are imbalanced.
2. Why do we need to deal with it:
The data is heavily biased to majority data which makes models overfitting on one class, although the accuracy might look good.
3. How to deal with imbalanced data:
To make our dataset balanced there are two ways to do so:
1) Under-sampling: Remove samples from over-represented classes; use this if you have a huge dataset
2) Over-sampling: Add more samples from under-represented classes; use this if you have a small dataset
Method:
SMOTE (Synthetic Minority Over-sampling Technique): It creates synthetic samples of the minority class. We use "imblearn" python package to over-sample the minority classes.
4. How to evaluate?
Accuracy as evaluation metrics makes sense only if the class labels are uniformly distributed. In the case of imbalanced classes confusion-matrix is a good technique to summarizing the performance of a classification algorithm.
For the outcome which has more than two classes, the number between the actual labels and predicted labels is much lower than others. After implementing the method, if accuracy and the number from a confusion matrix
increase, it works.
reference: https://towardsdatascience.com/m ... ata-set-29f6a177c1n
A/B Testing: AB TESTING 简单总结 https://towardsdatascience.com/a ... course-9ecc32dedbb1
看了udacity的课程,感觉自己金鱼记忆,忘记完了。。只能先硬记一下熟悉熟悉
面试关:
投简历状况:海投觉得没效果,找人内推又有点麻烦。。多在地里看看,去indeed啥的一键投吧。但是需要多积累。
Behavior Questions: 练习了三个bq: tell me about yourself/ the project you are proud of / the strength you have.
看了amazon的Leadership Principles, 找了地里的总结。。没见过这么变态的bq。。。
|
|