Why Your Big Data Project Will Fail, and How to Avoid It
•Download as PPTX, PDF•
0 likes•481 views
Why Classification of your Big Data matters with examples of failed classification or interpretation of datum can have bad consequences.
Recommended
More Related Content
Similar to Why Your Big Data Project Will Fail, and How to Avoid It
Similar to Why Your Big Data Project Will Fail, and How to Avoid It (20)
Recently uploaded
Recently uploaded (20)
Why Your Big Data Project Will Fail, and How to Avoid It
- 1. Big Data and Classification Paul Balas Content Architect 303Computing
- 2. A Dystopian Future George Orwell feared those who would deprive us of information. He feared the truth would be concealed from us.
- 3. He never imagined “Big Data” A glut of information that would conceal truth unintentionally
- 4. In 9.5 minutes or less… Convince you that without classification BIG DATA FAILS
- 6. Methods for Classification Data Mining Classification Machine-driven Taxonomies Human-driven
- 7. Classification Helps! • Group information by common attributes • Easily compare similarities and differences
- 8. People Classify (most) All classification is done by humans at some point in the life of a datum Not Machines Not Algorithms
- 9. But Deep Learning is Changing This • Machines can now ‘recognize’ complex objects without supervision
- 10. Without Classification, finding information is like finding a needle in a haystack…
- 11. Or, mistaking the haystack for a pile of needles
- 12. With Big Data, the haystack is huge
- 13. People don’t always agree with rules of the game for example Super Bowl XL Scott Steinmann
- 14. A Quiz for you… On the next slide, I want you to tell me what these four types of data have in common Raise your hand when you get the answer… (don’t worry, I won’t call on anyone)
- 15. “A computer would deserve to be called intelligent if it could deceive a human into believing that it was human.”
- 16. Did you get it right? Alan Turing The more data types we have The harder the classification
- 17. Classification Cracked The Enigma Code 158,962,555,217,826,360,000 possibilities Turing used Classification of the data to narrow the problem set 1st A letter can never be itself 2nd Known Phrases - The weather report
- 18. Without Classification There is no Correlation Without Correlation We are all out of jobs!
- 19. The ‘Classification Food Chain’ Classification shapes data Shaped data enables data quality Data Quality delivers confidence in results
- 20. Bad Classification Has Bad Consequences Elections are won Shuttles explode Financial Markets Meltdown
- 21. If you want to be confident in your Big Data results… Invest in your classifications as they are critical to your success!
Editor's Notes
- Big Data and Classification – why more than ever, classification and good data architecture is critical to providing confidence in analytical outcomes for your big data project. Paul Balas 25 + years of data architecture and data-centric implementations from sourcing all types of data, to sourcing, mastering, modeling, and sharing Implemented numerous content architectures for fortune 500 companies and the Kingdom of Saudi Arabia – KAPSARC Chair of the Big Data in Denver LinkedIn Group.
- He never imagined a world with so much data that information would be obscured simply by the volume and variety of data.
- If you can’t categorize your data, you can’t analyze it. If you aren’t performing data profiling on your big data as a first step to your analysis, you’ve already failed.
- The earliest known system of classification is that of Aristotle, who attempted in the 4th cent. B.C. to group animals according to such criteria as mode of reproduction and possession or lack of red blood. Aristotle's pupil Theophrastus classified plants according to their uses and methods of cultivation. Little interest was shown in classification until the 17th and 18th cent., when botanists and zoologists began to devise the modern scheme of categories. The designation of groups was based almost entirely on superficial anatomical resemblances.
- Machine driven classification can assist in human analysis and refinement of classification systems, but as it stands, without human context, machine driven classification is limited.
- Well-known classification systems such as GAAP-based accounting or plant taxonomies provide a common language that is widely accepted and therefor trusted. Common classification systems facilitate understanding and knowledge sharing.
- A new Big Data phenomenon is the ‘Data Lake’. I like to call it the ‘Data Swamp’ as the information added to the lake is useless until it’s classified. The excitement around Hadoop and other NoSQL technologies is it allows you to defer classfication, cleansing, and standardization post-load and on the fly, thus making the ingestion process and certain types of analytical workloads much faster
- Billions of dollars and tens of thousands of person-years effort has been spent on search technologies all with the focus of classifying data on-the-fly to help people locate precise information. Most of this effort has been driven by the internet search engines and firms trying to capitalize on e-commerce.
- Bad categorization of a population has the effect of completely misleading results and creating controversy NASA: Ninety-seven percent of climate scientists agree that climate-warming trends over the past century are very likely due to human activities,1and most of the leading scientific organizations worldwide have issued public statements endorsing this position. The following is a partial list of these organizations, along with links to their published statements and a selection of related resources. The Wall Street Journal The Myth of the Climate Change '97%' What is the origin of the false belief—constantly repeated—that almost all scientists agree about global warming? By JOSEPH BAST And ROY SPENCER Ms. Oreskes's definition of consensus covered "man-made" but left out "dangerous"—and scores of articles by prominent scientists such as Richard Lindzen, John Christy, Sherwood Idso and Patrick Michaels, who question the consensus, were excluded. The methodology is also flawed. A study published earlier this year in Nature noted that abstracts of academic papers often contain claims that aren't substantiated in the papers.
- According to IBM, In 2015 Global Data Volume is about 8,000 exabytes Most of it is sensor and social media data By 2020 some predict a 5x growth to 40,000 exabytes
- Even though he was already ejected from the game, Scott Steinmann continues to argue with the umpires call. Why were there so many controversial calls in Super Bowl XL? Where the rules for each penalty applied fairly? The outcome of that game was hotly debated.
- What was easy for those of you who knew the answer, is exponentially difficult for machines. Each data type has to be parsed and a common taxonomy applied as metadata to the data itself, then correlated to find the commonalities in each data source.
- That is 158 Quintillion if you wanted to know.
- Was the chad in favor of Bush or Gore? The risk of the o-ring failure wasn’t correctly classified based on the temperatures it would encounter. Credit default swap risk wasn’t correctly categorized and risky financial decisions ensued.