articles_2014.md

2014

• Anscombe's quartet

  

• A modern guide to getting started with Data Science and Python

  • numpy -> pandas
  • matplotlib -> seaborn

2014/11/16

• 단어

  • 확증 편향: 신념과 일치하는 정보만 받아들이는 경향
  • 잘못된 인과관계의 오류: 단순 선후 관계 사건을 인과관계로 잘못 해석하는 것

• ipython

  • Reloading submodules in IPython

   %load_ext autoreload
   %autoreload 2

• python

  • sort a list of dictionaries by value of the dictionary

   from operator import itemgetter
   newlist = sorted(list_to_be_sorted, key=itemgetter('name'), reverse=True) 

  • Python Packages For Datamining

• Data Mining

  • 8 Skills You Need to Be a Data Scientist
  • 빅 데이터 자체는 아무 의미도 없다… 맥락과 해석 찾는 힘은 인문학에서 나와
  • 38 Seminal Articles Every Data Scientist Should Read

• etc

  • Color Hex Color Codes

2014/11/03

• 기타

  • 젠킨스(Jenkins)를 이용한 지속적 통합(CI:Continuous Integration)

  • Mac OS에서 java경로 찾는 방법

     /usr/libexec/java_home -v 1.7  # java 1.7 path
     /usr/libexec/java_home -v 1.6  # java 1.6 path

  • Cryptographic hash function

    • 해시 값으로부터 원래의 입력값과의 관계를 찾기 어려운 성질을 가지는 해쉬함수

2014/10/31

• hive
  • hive sampling
  • HiveQL: Data Definition
  • LanguageManual UDF

2014/10/28

• AOV = total sales revenue / total number of sales

  • Is There a Value to Average Order Value (AOV)?
  • 3 Optimization Tips for Increasing Your Average Order Value

• Information Gain vs. Gain Ratio

  • Information Gain Ratio
  • Iterative Dichotomization
  • Information gain and information gain ratio: Do I have to pick just one?
  • Information Gain Versus Gain Ratio: A Study of Split Method Biases
  • Decision-Tree Learning

• Data Analysis

  • Ask a Data Scientist: The Bias vs. Variance Tradeoff
  • Interpreting the Pearson coefficient

• 기타 링크

  • python 7.1.3. Format String Syntax
  • Setup a remote IPython notebook server

2014/10/24

• KDE - Kernel Density Estimation

  • Wikipedia
  • Google Scholar
  • Kernel Density Estimation in Python
  • seaborn.kdeplot

• One-hot encoding

  • One-hot/One-of-k Data Encoder for Machine Learning
  • One Hot Encoding for Machine learning - stackoverflow
  • How can i vectorize list using sklearn DictVectorizer
  • sklearn.feature_extraction.DictVectorizer

2014/10/23 오랜만에 다시

• 디지털 광고 단가, ‘클릭’에서 ‘시간’으로 바뀌나

  • 유저에게 광고가 노출된 시간으로
  • 누구에게 유리할것인가?
  • 광고가 유저에게 오래동안 노출되면 annoying이 될것 같은데..

• 친절마저 상품이 된 시대, 혁명은 없다

  • 공유경제를 안좋게 보자면, 개인의 모든것을 상품화 시켜서 삶자체를 상품으로 만든다는 이야기.
  • 남는 시간, 남는 공간, 남는 친절을 모두 상품화해서 판매하는 개인.
  • 하지만 당분간 이쪽으로 새롭게 나오는 서비스는 계속해서 늘어날것 같다.

• 네이버 라인의 부가서비스 파헤쳐보기

  1. 라인페이 (결제, 예정)
  2. 라인와우 (배달서비스, 예정)
  3. 라인택시 (콜택시, 예정)
  4. 라인맵스 (지도서비스, 예정)
  5. 라인앳 (상업용 계정, 예정)
  6. 라인 유료 공식계정 (연예인용 광고계정, 예정)
  7. 라인게임 (게임, 예정)
  8. 라인뮤직 (음원, 예정)
  9. 라인블로그 (블로그 플랫폼, 예정)
  10. 라인비즈니스커넥트 (비즈니스 메신저)
  11. 라인몰 (전자상거래)
  12. 라인스토어 (라인 관련 쇼핑몰)
  13. 라인데코 (스마트폰 꾸미기)
  14. 기타

2014/06/01 Read Later 정리

• 시각화

  • http://ppss.kr/archives/20564
  • https://www.instapaper.com/read/477171569
  • https://www.instapaper.com/read/477819934

• CrossDomain

  • http://stove99.tistory.com/10
  • http://www.tuicool.com/articles/U7VVBf
  • http://net4all.tistory.com/6
  • http://warmz.tistory.com/739
  • http://webtn.tistory.com/entry/javascriptCross-Domain-%EA%B9%A8%EB%B6%80%EC%88%98%EA%B8%B0
  • http://blog.naver.com/sheis0/120208400734

• ReactiveMongo, Play & AngularJS

• OpenSSL버그 Heartbleed

• 확률적 자료구조를 이용한 추정 - 유일한 원소 개수(Cardinality) 추정과 HyperLogLog

• How to shuffle songs?

• Naive Bayes vs. Logistic Regression

• The Data Analytics - Free Book

  1. Communication skills are underrated If you can't present your analysis into digestible concepts for your CEO to understand, your analysis is only useful to yourself.
  2. The biggest challenge for a data analyst isn't modeling, it's cleaning and collecting Data analysts spend most of their time collecting and cleaning the data required for analysis. Answering questions like "where do you collect the data?", "how do you collect the data?", and "how should you clean the data?", require much more time than the actual analysis itself.
  3. A Data Scientist is better at statistics than a software engineer and better at software engineering than a statistician The greatest difference between a data scientist and a data analyst is the understanding of computer science and conducting analysis with data at scale. Data scientists only need a basic competency in statistics and computer science and not all are Ph.Ds. New tools are empowering more people to do data science.
  4. Do your own projects to break into the industry. The truth is, even in a quantitative major you are not taught what you need to know to work in data analytics. There is a learning gap between academia and industry that is best filled by doing projects. Find some sports statistics and do your own analysis. Learn R so that you can complete this analysis, not just to learn R itself. Also try Kaggle.
  5. Statistics > Programming. The development of tools and popularity of programmers has caused black box statistical analysis usage. Understanding selection bias vs. sampling bias and the underlying assumptions to which statistical functions are built on will make your opinions matter and your work invaluable.
  6. The most important skill is being able to ask the right questions. The power of data analytics is in taking open response questions and framing them to be multiple choice. Therefore if you have the ability to filter a million questions into options A through D, you are a data scientist for hire.

• 온라인 뉴스

  • NSA Guardian
  • Snow Fall
  • NYT 사진 아카이빙

• Python

  • Data streaming in Python: generators, iterators, iterables
  • Unit Testing with Python
  • Surprising Python

• AngularJS

  • https://angularjs.org/
  • AngularJS 와 RequireJS 를 활용한 대규모 웹 어플리케이션 개발
  • Bite-sized web development training with AngularJS
  • Use AngularJS to Power Your Web Application
  • http://mgcrea.github.io/angular-strap/

• A/B Test, Multi-armed Bandit

  • Agile A/B testing with Bayesian Statistics and Python
  • 다중 슬롯머신 실험
  • Introduction to Bandits: Algorithms and Theory
  • Python Multi-armed Bandits (and Beer!)

• Bayesian

  • Bayesian First Aid
  • Bayesian First Aid: Binomial Test
  • Bayesian First Aid: Two Sample t-test
  • Bayesian estimation supersedes the t test.

• Courses

  • StatLearning: Statistical Learning

  • Data Science in Python

  • Getting Started with Python for Data Scientists

  • Python Tools for Machine Learning

  • open-source-data-science-masters-curriculum

  • One Page R: A Survival Guide to Data Science with R

  • ggplot2: Cheatsheet for Visualizing Distributions

  • Time series components

  • http://tryr.codeschool.com/

  • An Introduction to Deep Learning: From Perceptrons to Deep Networks

  • Hadoop 101: The Most Important Terms, Explained

  • 9 Free Books for Learning Data Mining & Data Analysis

• Machine Learning 관련 Services

  • http://www.wise.io/
  • http://www.tellapart.com/

2014/04/04

• This is Digital Marketing: From Ad Men to Math Men | Mediative

• 국립현대미술관 서울관 <쉬린 네샤트_Shirin Neshat>

• 재미있는 인터넷 법칙들

    커닝햄 법칙 (Cunningham's Law)
    The best way to get the right answer on the Internet is not to ask a question, it’s to post the wrong answer.
    인터넷에서 정답을 알아내는 최고의 방법은, 질문을 하는 것이 아니라 잘못된 답을 올려놓는 것이다.
    http://meta.wikimedia.org/wiki/Cunningham's_Law
  
    영원한 9월(Eternal September)
    “인터넷 개통 축하드립니다.”라는 의미로 쓰는 듯.
    http://en.wikipedia.org/wiki/Eternal_September
  
    스트라이젠드 효과 (Streisand effect)
    인터넷 확산을 막으려고 법적 대응을 하다가 그것 때문에 역으로 더 유명해지는 현상.
    http://en.wikipedia.org/wiki/Streisand_effect
  

2014/01/01

• sed one line: http://sed.sourceforge.net/sed1line.txt
• awk one line: http://www.pement.org/awk/awk1line.txt