Machine Learning Model Validation (Aijun Zhang 2024).pdf
Sonamine casual connect july 2011 beyond metrics - predictives
1. Nick Lim Casual Connect CEO Online Social Games Track Sonamine July 21, 2011 (Revised: DRS, July 28) Beyond Metrics: Predictives With lessons and perspectives from other industries
3 weeks’ worth of CapitalOne credit-card offers! This diagram is taken from a recent blog post on consumerist.com http://consumerist.com/2011/07/capital-one-sends-you-over-20-credit-card-offers-in-3-weeks.html Predictives make sure you don’t get sent mail you don’t want.
Metrics = measuring things. After you measure things, you have to tell people what you measured. Then you can ask questions about what you measured, and you can be notified regarding measured things. This diagram is taken from the popular book Competing on Analytics by Tom Davenport and Jeanne Harris. http://www.amazon.com/Competing-Analytics-New-Science-Winning/dp/1422103323
If that example top decile has 4X lift, it will therefore have 40% of the total player-conversions. For more on lift, including a curve showing the front-loading effect (the cumulative gains) of predictions, and a lift curve: http://www2.cs.uregina.ca/~dbd/cs831/notes/lift_chart/lift_chart.html