In anticipation of her upcoming conference presentation, Crowd-Sourcing and Quality: How To Get The Best Out of Hand-Tagged Training Data for Machine Learning Models at Predictive Analytics World for Business New York, Oct 29-Nov 2, 2017, we asked Leslie Barrett, Search Products at Bloomberg L.P., a few questions about her work in predictive analytics.
Q: In your work with predictive analytics, what behavior or outcome do your models predict?
A: Our work with predictive analytics involves mostly text-based supervised learning models.
Q: How does predictive analytics deliver value at your organization – what is one specific way in which it actively drives decisions or operations?
A: We have developed sophisticated models that predict prices for illiquid and infrequently traded commodities – our Liquidity Assessment Tool (LQA). These models enable institutional investors to estimate fair prices before they make a trade.
Q: Can you describe a quantitative result, such as the predictive lift of your model or the ROI of an analytics initiative?
A: After we developed new deep learning techniques for table recognition and segmentation, we improved the precision of our automated extraction pipeline to above 90%.
Q: What surprising discovery or insight have you unearthed in your data?
A: Over the course of my experience in machine learning, what most surprises me is how biased almost all samples are. It is very difficult to randomize the population of all values for all features simultaneously. One needs to know the tradeoffs of overfitting up front in most cases.
Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.
A: In my talk, “Crowdsourcing & Quality: Getting the best out of hand-tagged training data for machine learning models,” I aim to provide the audience with new insights – and, I hope, greater confidence – in embarking on supervised learning tasks that involve hand-annotated training data.
Don’t miss Leslie’s conference presentation, Crowd-Sourcing and Quality: How To Get The Best Out of Hand-Tagged Training Data for Machine Learning Models, on Monday, October 30, 2017 at 10:30 to 11:15 am at Predictive Analytics World New York. Click here to register to attend. Use Code PATIMES for 15% off current prices (excludes workshops).
By: Eric Siegel, Founder, Predictive Analytics World
Eric Siegel is the founder of Predictive Analytics World (www.pawcon.com) — the leading cross-vendor conference series consisting of 10 annual events in New York, Chicago, San Francisco, Washington DC, London, and Berlin — and the author of the award-winning book, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die – Revised and Updated Edition, (Wiley, 2016).