Displaying results 1 - 2 of 2
-
Rapid classification of autism for public health surveillance
Content Type: Webinar
This presentation given August 3, 2017 describes work toward applying machine learning methods to CDC’s autism surveillance program. CDC’s population-based autism surveillance is labor-intensive and costly, as it requires clinicians to manually… read more… Forests1 • Ensemble classifier, 10,000 trees initially Training Data: 2008 Georgia ADDM site • 1,162 children (601 … pandas) 1. Breiman, 2001 Absent Random forests: training one tree PresentAbsent “eye contact” “autism” … -
Human-learned lessons about machine learning in public health surveillance
Content Type: Webinar
Presented December 13, 2018. For public health surveillance, is machine learning worth the effort? What methods are relevant? Do you need special hardware? This talk was motivated by these and other questions asked by ISDS members. It will focus… read more… What methods are relevant? [cont'd] A good baseline: Train a model using Random Forests, Boosted Trees, or … sophisticated methods may perform. • A nice discussion of training a single model for multiple outcomes vs discrete … • For autism classification, we got similar performance training on 1162 observations vs ~3000. • More complex …