Causal Models for Software Cost Prediction & Control
October 2019 • Poster
Michael D. Konrad, Robert W. Stoddard, William Nichols, David Zubrow
This poster provides an update on ongoing research to use causal learning in software cost prediction.
Publisher:
Software Engineering Institute
Abstract
To reduce costs, the causes of an outcome (good or bad) need to be considered. Correlations are insufficient due to Simpson’s Paradox. Causal learning identifies when factors such as team membership explain away (or mediate) correlations, and it works for much more complicated datasets too.