Introduction to Data Science Section 1 Data Matters 2015 Sponsored by the Odum Institute, RENCI, and NCDS Thomas M. Carsey 1

Course Materials I used many sources in preparing for this course: – Practical Data Science using R by Zumel and Mount – – Data Mining with R: Learning with Case Studies, by Torgo – – An Introduction to Data Science, Version 3, by Stanton – – Monte Carlo Simulation and Resampling Methods for Social Science, by Carsey and Harden – &subject=J00&sortBy=defaultPubDate%20desc&fs=1#tabview=title &subject=J00&sortBy=defaultPubDate%20desc&fs=1#tabview=title – Machine Learning with R by Lantz – 2

Additional Materials A Simple Introduction to Data Science, by Burlingame and Nielsen simple_introduction_to_data_science simple_introduction_to_data_science Ethics of Big Data, by Davis Privacy and Big Data, by Craig and Ludloff Doing Data Science: Straight Talk from the Frontline, by O’Neil and Schutt 3

Learning R Lots of places to learn more about R – All of the sources on the first slide have R code available – Comprehensive R Archive Network (CRAN) – – Springer Textbooks Use R! Series – – Online search tool Rseek – – The RStudio site – – The Odum Institute’s online course – 4

What is Data Science? 5

What words come to mind when you think of Data Science? What experience do you have with Data Science? Why are you taking an Introduction to Data Science Class? 6

The Data Science Revolution Data science is exploding in importance and the attention it receives. It’s hard to sort through the substance and the hype. There is real value in data science, but you should have a purpose or goal in mind first. 7

8

The Roots of Data Science Simple observation and recording those observations dates back to the most ancient civilizations – The Greeks were the first western civilization to adopt observation and measurement Some call Aristotle the first empirical scientist – Muslim scholars between the 10 th and 14 th centuries developed experimentation (Haytham) – Roger Bacon ( ) promoted inductive reasoning (inference) – Descartes ( ) shifted focus to deductive reasoning. 9

What is Data Science? “How Companies Learn Your Secrets” NYT, by Charles Duhigg, February 16, ing-habits.html?pagewanted=1&_r=2&hp& ing-habits.html?pagewanted=1&_r=2&hp& 10

What did Target Do? Mining of data on shopping patterns – Specific products purchased – Combination of products purchased – Combined with demographic and other data Psychology and neuroscience – Habits: Cue-routine-reward When are habits open to change? 11

Lessons from Target Yes, Data Science is about mining data There are deeper theoretical issues involved in understanding what you find Left out of that long article are most of the critical steps that precede the analysis In short, Data Science > data mining 12

Definition of Data Science There are many, but most say data science is: – Broad – broader than any one existing discipline – Interdisciplinary: Computer Science, Statistics, Information Science, databases, mathematics Also substantive domains (environmental science, sociology, public health, etc.) – Applied focus on extracting knowledge from data to inform decision making. – Focuses on the skills needed to collect, manage, store, distribute, analyze, visualize, and reuse data. There are many visual representations of Data Science 13

Some definitions link computational, statistical, and substantive expertise 14

Other definitions focus more on technical skills alone 15

Still other definitions are so broad as to include nearly everything 16

There are many “Word Cloud” representations of Data Science as well 17

18

19

Definition of Data Science The field is immature, cluttered by hype, unfocused. But, key features should include: – Data across its lifecycle – Interdisciplinary skills – Substantive knowledge 20

Defining Some Terms 21

MapReduce and Hadoop – Designed to process large operations quickly – Distributes the problem across multiple servers – The Map part filters and sorts data into bins or queues based on some share characteristic – The Reduce part then executes some operation on each bin of data. – Results are then reassembled – It is like parallel processes, but distributed across servers rather than just processors – Scalable and has a fault tolerance – Hadoop is an open-source version 22

More on MapReduce Pig – Software platform used for creating MapRedce programs used by Hadoop. Hive – A date warehouse infrastructure built on top of Hadoop. Used to query, summarize, or analyzed data. 23

Database Management SQL – Structured Query Language – A programming language designed for management of relational databases MySQL – Open source implementation of an SQL-like system for management of relational databases (used by Wikipedia, Google, Facebook, Twitter, Flickr, YouTube) NoSQL – (Not Only SQL) – Used for databases where the data is in some form other than tabular relations like those used in relational databases – Cassandra (Apache) Distributed database management with not single node of failure Scalable with no down time 24

25

Cloud Computing Standard client-server model where computing operations don’t happen on the local (desktop) machine. What’s new? Virtualization. You are not connecting to a specific server. – Servers are virtual – One server can run multiple virtual machines – One virtual machine can use multiple servers This makes the “machine” scalable, moveable, configurable. Allows selling software, platforms, and even computing infrastructure as a “service” 26

27

Data Mining/Machine Learning Machine learning uses computer algorithms to get a machine to learn and adapt to new information. Data Mining more explicitly focuses on discovering patterns or structure in a given set of data. Often used as synonyms by non-experts without much loss of information. 28

29

Web Scraping This is a process of collecting information from websites and then organizing it for some sort of analysis. Scraping is just about getting the data; the analysis comes later. 30

Programming Tools – R – Statistical programming (object oriented, scripting language) – Python – a scripting programming language that supports object-oriented programming, structured programming, functional programming – SQL – Relational Database – SAS – General purpose data analysis software – Julia – Faster than R and more scalable than Python – Kafka and Storm – Used for real-time streaming analysis 31