Rice University Bayesian Methods for Learning Analytics with Applications Andrew E. Waters A
Presentation by Andrew Jeavons, Survey Analytics
-
Upload
northwest-mra -
Category
Business
-
view
558 -
download
1
description
Transcript of Presentation by Andrew Jeavons, Survey Analytics
![Page 1: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/1.jpg)
Big Data
Andrew JeavonsPresident
Survey Analytics LLC
![Page 2: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/2.jpg)
Overview•Coming out of the data desert.
•What do we mean by Big Data ?
•Behavior, Big Data and Market Research.
•Get a model to get ahead.....
![Page 3: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/3.jpg)
Data Desert
•Data was scarce..
•The hunt for completes.
•Sampling ?
•Limited data.
•Limited potential...
![Page 4: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/4.jpg)
The Deluge•The Data Tsunami.
•What we get..
•What we don’t get...
![Page 5: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/5.jpg)
Forms of Data
•Metadata, Paradata, Data
•Big Data and Prediction of behavior.
•The central dogma:
•Big Data is Marxist.
•Black Swans - Taleb’s cautionary tale...
![Page 6: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/6.jpg)
Machine Learning
•What is meant by this ?
•Learning and pattern recognition.
![Page 7: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/7.jpg)
Nerves and Genes
•Neural Networks
•Genetic Algorithms
•“Agnostic” algorithms...
![Page 8: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/8.jpg)
Neural NetworksOriginally sought to reproduce the way the nervous system works.
Had the ability to learn complex relationships with any assumptions.
Can learn complex non linear relationships.
![Page 9: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/9.jpg)
Back Propogation
![Page 10: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/10.jpg)
Genetic Algorithms
![Page 11: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/11.jpg)
Biological metaphor: encode problems as chromosomes.Fitness function.Evolve different mapping functions via principles of evolution.Cross over and mutation (note: mutation rates not as powerful as you may think...)
MR example: used GA’s to evolve panel sample balancing - rim balancing.
![Page 12: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/12.jpg)
Social Media Data
• We have a lot of text...that we didn’t have before...
• Text has its own set of problems..unlike Gore Vidal we don’t speak in sentences.
• “#itsucks phat”
• Txt messages, social media messages.
• Parsing and tokenizing
![Page 13: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/13.jpg)
Sentiment Analysis
•One of the frontiers for MR and Big Data
•Can we tell the sentiment behind the comment ?
•Text is messy because language is messy.
•Grammatical vs Stochastic
![Page 14: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/14.jpg)
Surveys, Market Research and Big Data
•Fusion of data
•Survey and behavioral
•“Whole respondent” view
•Surveys can be “big data”
![Page 15: Presentation by Andrew Jeavons, Survey Analytics](https://reader033.fdocuments.us/reader033/viewer/2022061111/5455468daf7959755d8b4774/html5/thumbnails/15.jpg)
Prediction and Market Research
•Statisticians are the new “rock stars” (Eric Schmidt)
•MR has to embrace prediction and realize the limits of explanation. It’s great to know why, better to know what’s next ?
•Theories of respondents and consumers...