Geo Analytics Tutorial - Where 2.0 2011

Geo Analytics TutorialPete SkomorochSr. Data Scientist - LinkedIn (@peteskomoroch)#geoanalytics** Hadoop Intro slides from Kevin Weil, Twitter

Topics‣ Data Science & Geo Analytics‣ Useful Geo tools and Datasets‣ Hadoop, Pig, and Big Data‣ Cleaning Location Data with Mechanical Turk‣ Spatial Tweet Analytics with Hadoop & Python‣ Using Social Data to Understand Cities‣ Q&A

Analytics & Data are Hot Topics

Data Exhaust My Delicious Tags

Data Science * http://www.drewconway.com/zia/?p=2378

Data Visualization ‣ http://www.dataspora.com/blog/

Spatial Analysis Map by Dr. John Snow of London, showing clusters of cholera cases in the 1854 Broad Street cholera outbreak. This was one of the first uses of map-based spatial analysis.

Spatial Analysis• Spatial regression - estimate dependencies between variables• Gravity models - estimate the flow of people, material, or information between locations• Spatial interpolation - estimate variables at unobserved locations based on other measured values• Simulation - use models and data to predict spatial phenomena

Life Span & Food by Zip Code* http://zev.lacounty.gov/news/health/death-by-zip-code* http://www.verysmallarray.com/?p=975

Where Americans Are Moving (IRS Data) ‣ (Jon Bruner) http://jebruner.com/2010/06/the-migration-map/

Facebook Connectivity (Pete Warden)* http://petewarden.typepad.com/searchbrowser/2010/02/how-to-split-up-the-us.html

Useful Geo Tools•R, Matlab, SciPy, Commercial Geo Software•R Spatial Pkgs http://cran.r-project.org/web/views/Spatial.html•Hadoop, Amazon EC2, Mechanical Turk•Data Science Toolkit: http://www.datasciencetoolkit.org/•80% of effort is often in cleaning and processing data

DataScienceToolkit.org•Runs on VM or Amazon EC2•Street Address to Coordinates•Coordinates to Political Areas•Geodict (text extraction)•IP Address to Coordinates•New UK release on Github

Resources for location data• SimpleGeo• Factual• Geonames• Infochimps• Data.gov• DataWrangling.com

Hadoop: Motivation •We want to crunch 1TB of Twitter stream data and understand spatial patterns in Tweets •Data collected from the Twitter “Garden Hose” API last Spring

Data is Getting Big‣ NYSE: 1 TB/day‣ Facebook: 20+ TB compressed/day‣ CERN/LHC: 40 TB/day (15 PB/year!)‣ And growth is accelerating‣ Need multiple machines, horizontal scalability

Hadoop‣ Distributed file system (hard to store a PB)‣ Fault-tolerant, handles replication, node failure, etc‣ MapReduce-based parallel computation (even harder to process a PB)‣ Generic key-value based computation interface allows for wide applicability‣ Open source, top-level Apache project‣ Scalable: Y! has a 4000-node cluster‣ Powerful: sorted a TB of random integers in 62 seconds

MapReduce?cat file | grep geo | sort | uniq -c > ‣ Challenge: how many tweets peroutput county, given tweets table? ‣ Input: key=row, value=tweet info ‣ Map: output key=county, value=1 ‣ Shuffle: sort by county ‣ Reduce: for each county, sum ‣ Output: county, tweet count ‣ With 2x machines, runs close to 2x faster.

But...‣ Analysis typically done in Java‣ Single-input, two-stage data flow is rigid‣ Projections, filters: custom code‣ Joins: lengthy, error-prone‣ n-stage jobs: Hard to manage‣ Prototyping/exploration requires ‣ analytics in Eclipse? compilation ur doin it wrong...

Enter Pig ‣ High level language ‣ Transformations on sets of records ‣ Process data one step at a time ‣ Easier than SQL?

Why Pig?‣ Because I bet you can read the following script.

A Real Pig Script‣ Now, just for fun... the same calculation in vanilla Hadoop MapReduce.

No, seriously.

Pig Simplifies Analysis‣ The Pig version is:‣ 5% of the code, 5% of the time‣ Within 50% of the execution time.‣ Pig Geo: ‣ Programmable: fuzzy matching, custom filtering ‣ Easily link multiple datasets, regardless of size/structure ‣ Iterative, quick

A Real Example‣ Fire up your Elastic MapReduce Cluster. ‣ ... or follow along at http://bit.ly/whereanalytics‣ I used Twitter’s streaming API to store some tweets‣ Simplest thing: group by location and count with Pig ‣ http://bit.ly/where20pig‣ Here comes some code!

tweets = LOAD s3://where20demo/sample-tweets as ( user_screen_name:chararray, tweet_id:chararray, ... user_friends_count:int, user_statuses_count:int, user_location:chararray, user_lang:chararray, user_time_zone:chararray, place_id:chararray, ...);

tweets_with_location = FILTER tweets BY user_location !=NULL;

normalized_locations = FOREACH tweets_with_locationGENERATE LOWER(user_location) as user_location;

grouped_tweets = GROUP normalized_locations BYuser_location PARALLEL 10;

location_counts = FOREACH grouped_tweets GENERATE $0 aslocation, SIZE($1) as user_count;

sorted_counts = ORDER location_counts BY user_count DESC;

STORE sorted_counts INTO global_location_tweets;

hadoop@ip-10-160-113-142:~$ hadoop dfs -cat /global_location_counts/part* | head -30brasil 37985indonesia 33777brazil 22432london 17294usa 14564são paulo 14238new york 13420tokyo 10967singapore 10225rio de janeiro 10135los angeles 9934california 9386chicago 9155uk 9095jakarta 9086germany 8741canada 8201 7696 7121jakarta, indonesia 6480nyc 6456new york, ny 6331

Neat, but... ‣ Wow, that data is messy! ‣ brasil, brazil at #1 and #3 ‣ new york, nyc, and new york ny all in the top 30 ‣ Mechanical Turk to the rescue...

Code examples we’ll cover are on Github

You can run them on Elastic MapReduce

Cleaning Twitter Profile Location Names Filter Exact MatchesExtract Top Tweet Locations Clean with MTurk Aggregate Context with Hadoop

We will map locations to GeoNames IDs

Start with Location Exact Matches

Use Mechanical Turk to improve results

Workers do simple tasks for a few cents

We constructed the following task

Workers used a Geonames search tool

Location search tool code is on Github

Preparing Data to send to MTurk

We use consensus answers from workers

Processing MTurk Output

Tokenizing and Cleaning Tweet Text‣ Extract Tweet topics with Hadoop + Python + NLTK + Wikipedia

Build Phrase Dictionary with Wikipedia

Streaming Tweet Parser (Python + NLTK)

Parse Tweets and Join to Wikipedia (Pig)

Aggregate by US County for Analysis

Clean Data => Thematic US County Map

Twitter users by county in our sample

“Lady Gaga” Tweets

“Tea Party” Tweets

“Dallas” Tweets

“Stephen Colbert” Tweets

LinkedIn Skills

Skills in the Design Industry

Exploring the Spatial Distribution of Skills

People with “Ship Building” Skills

What is the Skill profile of a given city?

Expertise correlated with Santa Clara, CA

Expertise correlated with Los Angeles

Expertise correlated with Washington, DC

Yuba City, CA has 21.3% Unemployment 21.3

Ames, Iowa has 4.7% Unemployment 21.3

Questions? Follow me at twitter.com/peteskomoroch datawrangling.com

http://where2conf.com	306
http://blog.bookcold.me	273
http://a-ahad.blogspot.com	211
http://whereconf.com	83
http://bookcold.me	31
http://bookcold.com	24
http://www.techgig.com	23
http://paper.li	22
http://a-ahad.blogspot.com.es	21
url_unknown	8
http://a-ahad.blogspot.com.au	7
http://a-ahad.blogspot.co.uk	6
http://s240757377.onlinehome.fr	5
http://webcache.googleusercontent.com	3
http://twitter.com	3
http://173.230.153.79	2
http://a-ahad.blogspot.ca	1
http://3scrowd.com	1
http://tweetedtimes.com	1
http://a-ahad.blogspot.com.ar	1
http://a-ahad.blogspot.com.br	1
http://a-ahad.blogspot.pt	1

Geo Analytics Tutorial - Where 2.0 2011

by Peter Skomoroch on Apr 19, 2011

Accessibility

Categories

Tags

Upload Details

Usage Rights

22 Embeds 1,034

Statistics

Geo Analytics Tutorial - Where 2.0 2011 — Presentation Transcript