Cutting Edge Predictive Analytics with Eric Siegel
3 likes2,186 views
Apache Spark empowers predictive analytics and machine learning by increasing the reach and potential. But, before jumping to new deployments, it’s critical we 1) get the analytics right and 2) not overlook less conspicuous business opportunities. In this keynote, Predictive Analytics World founder and “Predictive Analytics” author Eric Siegel ramps you up on a dangerous pitfall and a critical value proposition: – PITFALL: Avoiding BS predictive insights, i.e., “bad science,” spurious discoveries – OPPORTUNITY: Optimizing marketing persuasion by predicting the *influence* of marketing treatments, i.e., uplift modeling
Ad
More Related Content
What's hot (20)
Similar to Cutting Edge Predictive Analytics with Eric Siegel (20)
![[REPORT PREVIEW] The AI Maturity Playbook: Five Pillars of Enterprise Success](https://cdn.slidesharecdn.com/ss_thumbnails/ai-maturity-playbook-altimeter-preview-181211202659-thumbnail.jpg?width=560&fit=bounds)
Ad
More from Databricks (20)
Ad
Recently uploaded (20)
Cutting Edge Predictive Analytics with Eric Siegel
- 1. For more information and citations regarding the examples in this presentation, see the book "Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die" by Eric Siegel (http://www.thepredictionbook.com). Most of the various examples shown are covered in the book - so, for greater detail about each case study named, see its reference/citation - search by organization name within the book's Notes PDF, available online at http://www.PredictiveNotes.com. 0
- 2. With 10 events a year globally, Predictive Analytics World delivers vendor- neutral sessions across verticals such as banking, financial services, e-commerce, entertainment, government, healthcare, manufacturing, high technology, insurance, non-profits, publishing, and retail. Predictive Analytics World industry events include PAW Business, PAW Financial, PAW Government, PAW Healthcare, PAW Workforce, and PAW Manufacturing. Why bring together such a wide range of endeavors? No matter how you use predictive analytics, the story is the same: Predictively scoring customers, employees, students, voters, patients, equipment, and other organizational elements optimizes performance. Predictive analytics initiatives across industries leverage the same core predictive modeling technology, share similar project overhead and data requirements, and face common process challenges and analytical hurdles. Info: http://www.predictiveanalyticsworld.com
- 3. Attendees may receive a copy of the audiobook version at no charge by simply emailing me directly to request: [email protected] 2
- 4. Predictive analytics educational/infotainment geek rap video: http://www.PredictThis.org 3
- 5. 4
- 6. 5
- 7. Canadian Tire examples, from "What Does Your Credit-Card Company Know About You?" New York Times, May 12, 2009. http://www.nytimes.com/ 2009/05/17/magazine/17credit-t.html 6
- 8. Canadian Tire examples, from "What Does Your Credit-Card Company Know About You?" New York Times, May 12, 2009. http://www.nytimes.com/ 2009/05/17/magazine/17credit-t.html 7
- 9. People who “like” curly fries on Facebook are “more intelligent.” Reference for most examples in this presentation are in the Notes PDF for Eric Siegel's book, "Predictive Analytics." For each example's reference/citation, search by organization name within the book's Notes PDF, available at www.PredictiveNotes.com 8
- 10. Neighborhoods of San Francisco that exhibit higher rates of certain crimes also exhibit higher demand for Uber rides. 9
- 11. 10
- 13. 12
- 14. 13 A crummy predictive model delivers big value. It’s like a skunk with bling. Simple arithmetic shows the bottom line profit of direct mail, both in general and then improved by predictively targeting (and only contacting 25% of the list). The less simple part is how the predictive scores are generated for each individual in order to determine exactly who belongs in that 25%. For details on how this works, see Chapter 1 of the book "Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die" (http://www.thepredictionbook.com).
- 15. Put another way, predicting better than guessing is often sufficient to generate great value by rendering operations more efficient and effective. For details on how this works, see the Introduction and Chapter 1 of the book "Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die" (http:// www.thepredictionbook.com). 14
- 16. 15
- 17. This singular, universally applicable force improves every large-scale thing we do —how we build things, sell things, and prevent bad things from happening— because every function benefits from prediction. As its deployment takes hold across industries, we have moved beyond engineering the infrastructure that manages big data to implementing the science that taps its contents for more intelligent Has emerged as a commonplace business practice necessary to sustain competitive advantage. $6.5 billion within a couple years 84% of marketing leaders intend to increase the role of predictive analytics in marketing over the next 12 months (Forbes) U.S. shortage of analytics experts at 140,000 in the near-term LinkedIn’s first/second “Hottest Skills That Got People Hired” is “statistical analysis and data mining” 16
- 18. 17
- 19. 18 A crummy predictive model delivers big value. It’s like a skunk with bling.
- 20. 19
- 21. Does contacting the customer make them more likely to respond? MEDICAL: Will the patient survive if treated? "My headache went away!“ Proof of causality by example. Driving medical decisions with personalized medicine: selecting treatment, e.g., treating heart failure with betablockers Personalized medicine. Naturally, healthcare is where the term treatment originates. While one medical treatment may deliver better results on average than another, personalized medicine aims to decide which treatment is best suited for each patient, since a treatment that helps one patient could hurt another. For example, to drive beta-blocker treatment decisions for heart failure, researchers "use two independent data sets to construct a systematic, subject-specific treatment selection procedure." (Claggett et al 2011) Certain HIV treatment is shown more effective for younger children. (McKinney et al 1998) Cancer 20
- 22. a.k.a., net lift modeling and persuasion modeling 21
- 23. 22
- 24. US BANK EXAMPLE Business case: Direct mail for a home-equity line of credit Approach: Target campaign with an uplift model … to existing customers (i.e., a cross-sell marketing campaign) Resulting improvements over prior conventional analytical approach: Campaign ROI increased over 5 times previous campaigns (75% to 400%) Cut campaign costs by 40% Increase incremental cross-sell revenue by over 300% Decrease mailings to customers who would purchase whether contacted or not, and customers who would purchase only if not contacted. Sources: Radcliffe & Surry (2011), Tsai (2010), Patrick Surry (Pitney Bowes Business Insight), Michael Grundhoefer (US Bank) 23
- 26. 25
- 28. Two articles I wrote provide more details on the Obama campaign's use of predictive analytics (one a reprint of the pertinent section in my book): http://www.thefiscaltimes.com/Articles/2013/01/21/The-Real-Story- Behind-Obamas-Election-Victory http://bigthink.com/experts-corner/team-obama-mastered-the-science-of- mass-persuasion-and-won Then check out this video of a Predictive Analytics World presentation by one of the hands-on leads on this Obama project, which gets more technical: http://www.predictiveanalyticsworld.com/patimes/video-pinpointing- persuadables-convincing-right-customers-right-voters/ 27
- 29. Telenor: world’s 7th largest mobile operator - 159 million subscribers Improvements are relative to their existing best-practice retention models. Case study presented at Predictive Analytics World, February 2009, San Francisco. Case study and graph courtesy of Pitney Bowes Business Insight.
- 30. 29
- 31. Marcus Wohlsen, “San Francisco startup makes data science a sport.” The Seattle Times, April 14, 2012. 30
- 32. 31
- 33. 32
- 34. 33
- 35. 34
- 36. 35
- 37. 36
- 38. 37
- 39. 38
- 40. 39
- 41. 40
- 42. 41
- 43. 42
- 44. 43
- 45. 44
- 46. 45
- 47. 46
- 48. 47
- 49. “Last Week Tonight” with John Oliver, May 8, 2016: https://youtu.be/0Rnq1NpHdmw 48
- 50. 49
- 51. 50
- 52. 51
- 53. After 150 Years, the American Statistical Association Says No to p-values: http://www.kdnuggets.com/2016/03/asa-says-no-p-values.html "Even if something is published in the flagship journal of the leading association of research psychologists, there's no reason to believe it. The system of scientific publication is set up to encourage publication of spurious findings." - Andrew Gelman, professor of statistics and political science at Columbia University http://www.slate.com/articles/health_and_science/science/2013/07/ statistics_and_psychology_multiple_comparisons_give_spurious_results.2.html 52
- 54. Free white paper: www.predictiveanalyticsworld.com/signup-uplift-whitepaper.php More updated version thereof is Chapter 7 of “Predictive Analytics” (www.thepredictionbook.com) Or see the (much more) technical papers that chapter cites - see under Chapter 7 of the book's Notes PDF, available online at http://www.PredictiveNotes.com. See also: http://www.predictiveanalyticsworld.com/patimes/personalization-is-back-how-to-drive-influence-by-crunching-numbers/ 53
- 55. For more on vast search, see Chapter 3 of Predictive Analytics by Eric Siegel (www.thepredictionbook.com). For further citations on this topic, also see the final portion that chapters Notes (citations), available as a PDF online at www.PredictiveNotes.com. Also on a broadly-accessible level, see chapter seven, "Dead Fish Don't Read Minds," of "How Not to Be Wrong: The Power of Mathematical Thinking," by Jordan Ellenberg. Video: www.predictiveanalyticsworld.com/patimes/video-the-power-and-peril-of-predictive-analytics/ See also: http://www.predictiveanalyticsworld.com/patimes/breakthrough-avert-analytics-treacherous-pitfall/3366/ See also Elder Research's short, friendly video illustrating target shuffling: http://www.elderresearch.com/target-shuffling-video For a more technical treatment, in addition to Elder and Bullard's paper cited on the slide above, see "Handbook of Statistical Analysis and Data Mining Applications" by Nisbet et al - second edition only - Chapter 20, "SIGNIFICANCE vs LUCK in the age of MINING.” Coincidentally, I used "BS” in earlier versions of this talk before knowledge of similar vocabulary in that new edition of the “Handbook” and elsewhere J. Here’s a quote from that chapter: Replication is supposed to be a hallmark of scientific integrity. The concern over this lapse of ‘scientific accuracy/integrity’ has been growing for some time, especially over the past 3 years, such that the University of Washington (Seattle) is offering a new course for Spring 2017 titled "Calling Bullshit in the Age of Big Data" Among the Learning / Behavioral Objectives for this course is the following: ‘After taking this course, you will be able to provide a statistician or fellow scientist with a technical explanation of why a claim or conclusion is ‘in error’. Below are three links that go to this new course, including a syllabus: • http://callingbullshit.org/syllabus.html#Big; • http://www.recode.net/2017/2/19/14660236/big-data-bullshit-college-course-university- washington; • https://www.statnews.com/2017/02/17/science-fights-alternative-facts/ One more link: http://www.slate.com/articles/health_and_science/science/2016/01/amy_cuddy_s_power_pose_research_is_the_latest_example_of_scientific_overreach.html 54
- 56. 55
- 58. 57
- 59. 58