Getting a sensible model
Abstract
Judgment is involved in every step of scoring, starting with the big questions of what we even think we should be predicting. A huge issue is that predictions are not stable over time because the economy can be good or bad, people can lose jobs and not be at all like how they presented initially. A model that is best in one business frame is only mediocre in another. Modelers need to decide if they are scoring the probability of events that really threaten the enterprise or simply reduce profits. Models are usually measured by how well they order the consumers who fail vs. who succeed, illustrated with Gini, K–S, and similar statistics. I show some tricks for understanding the parameters ...
Get Credit Data and Scoring now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
