Thursday, September 19, 2013

2013 Predictive Analytics World Friday Notes

Gene Dodaro-data analytics for government oversight
  • Talkl about how to put gov't on a more sustainable path.
  • They advise congress on how to improve performance
  • Driven by mandates and requests from congress.
  • Fact-based organization-analysis hinges on reviewing, compiling, analyzing data
  • Made 380 recommendatin to reduce overlap.
  • 108 billion in improper payments (some agencies haven't even reported.
  • Health care is particularly worrisome.  5 to 8% of GDP in next few yaers.
  • Number of people 65 and older will double in next few years.
  • Overpaying people by tens of billions.

Panel on Working HOrizontally" Analytics as a bridget
  • Questiona sked about analytics just being repackaged as something we have always done.
  • Essentially agreed except that the repackaging simply helps to brand the usefulness of using data in making decisions.

Using Social Data for Public Sector Analytics
  • Will mayo and Rebecca Goolsby
  • Information moving rapidly.  
  • Shared example of tweet from San Francisco flight and news story didn't come in for 10 minutes beyond that tweet.
  • Used Social Media in public sector in response to Hurrican sandy (e.g. finding where a bunch of trees fell down)
  • Weather alerts via twitter where otherwise equipment is unavailable
  • Need to use consented data and that available in private sector.                                                                        


Big Data is not new-no such ting
  • Goal is to talk about what is data to your organization
  • Mid-2011 was when the google search term exponentially increased
  • No such thing as big data, but there is a lot of data out here
  • 234 milion e-mails per minute.
  • 2.5 quintillionb ytes of data created with 90% of worlds data created in last two years alone.
  • Data governance expert peter aiken estimates 80% of the data is not useful.
  • Is any of it valuable?
  • Big data can be defined as data sets that are too large for you to handle.
  • Origin of Big Data
    • Some credit John Mashey, chief scientist at silicon graphics in 1990's.
    • He was using the label for a range of issues, essentially that the boundaries of computing keep advancing.  
  • Nate silver labels big data as a fashionable word.  It is when we deny our role in the process of data driven predicitons that the odds of failure rise. (signal and the noise book)
  • Mark Madsen says big data isn't hype but it is being hyped and says tthe reality is that big data is about new models for data processing.
  • Sue Feldman says it is a set of technologies that solve complex information economically and includes volume, variety, and velocity.
  • The Hype
    • Gardner shows graph of hype cycle.  Big Data is two to five years from top of hype cycle.  
  • Business is about making decisions, data can help
  • We gotta do the hard work to figure out the value of business data.
  • Need to be willing to experiemnt and willing to be wrong.
  • Valuable data is not always big!!!!!!!!!!!!!!!!!!!!
  • Analytics can scale in a number of ways.  
  • BIG DATA INITIATIVES--frighten presenter
  • Technology looking for a problem.  Cart before the horse.
  • Need to first understand the problem before we shotgun some big data initiative.
  • Kudzoo????
  • Data Vs
    • Vacant is a new oen.  Available.  Availability of data.  people not wanting to share data.
    • Volume.  Big data implies volume.  Easiest one from analytical view to solve.  1 tterrabyte to 8 megabytes of useful information.  You can shrink that data.  Get rid of data rot.  
    • Velocity
    • Variety-biggest challenge.  Connecting data that wasn't designed to be conbined.  Fuzzy matching.  
    • Value-different colored bubble on slide.  He says its a diferent kind of question.
    • Veracity, what is the truth of the data and who says what the truth of that data is?
    • Vitality-is the system able to adjust around the data.
    • Variability-changing environment around us.  
  • CRISP-DM process for data mining.
    • Goal definition to business understanding to data understanding to data preparation to modeling to evaluation to deployment to knowledge application.  Operationalize it.  
  • Background-at least two components to any analytic architecture.
    • data storage (databases, dw, spreadsheets)
    • Analytical processing (dashboards, models, metrics, etc.)
  • They'd rather have the raw data, vice the aggregate.
  • Current standard is to combine storage with in-core analytics.  Limited interactionb etween storage and processing
  • In-database-effort to improve sall data by pushing analytics into data storage.  Avoids data transfer from databse to analytics. (relational databases, teradata, SAS, etc.).
  • Computational" 9 million records with 2000 attributes?  Big data?
  • What if you wanted to test 17 drug interactions on multiple morbidity outcomes?
  • It is about the data-treating data as an asset implies systems is designed to support
  • An asset is a resource controlled by org-data has a value.
  • You may have a lot of data
    • Value first through 'test and learn'
    • Data governance to maintain flexibility
    • Use technology to operationalize.
  • It's like an irritating fly buzzing around your head-----big data.
  • If they push back and say we need analytics
  • He says we should push back and ask what problem do you want to solve?
  • New OMB analytics guy said all our stuff neds to be structured. 
    • A lot of knowledge in tha tbusiness is in the people.  Not captured in the data.
    • problems are much more important to define. 
  • Apps
    • ENTERPRISE MINER FROM SAS SAVES A LOT OF TIME
    • HADOOP goes well for google in text analyzsis I believe.
    • Oracle does good things.
  • Don't need to be a data scientist to do big data
    • Rather have a heart surgeon to do surgeory on my heart.

David Jakubek.  Case Study-data to decisions building dta analytics capability in the department of defense.
  • Today is international talk like a pirate day.  Okay....
  • Told a really long joke.  Not worth the time imo
  •                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  `
Predictive Analytics in Medicare


  • asdf
  • asdf

No comments:

Post a Comment