Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Predictive Modeling topic
No spam. Unsubscribe anytime.
Newport seeks approval for random-forest model to prioritize likely lead service lines
Summary
Director Schultz said the city is using a random-forest predictive model that compiles building age, permits and other records to flag likely lead service lines; the city is seeking regulatory sign-off to reduce unnecessary inspections and letters.
At the workshop, Rob Schultz described the city's use of a 'random forest' predictive model to target inspections and prioritize replacement work rather than testing every parcel immediately.
"The 1 we're specifically using is random forest," Schultz said, explaining that the model ingests building permits, construction date and other records and combines multiple model runs to identify parcels with high confidence that a line is lead. He added that similar modeling approaches have been accepted in many states east of the Mississippi, but Rhode Island regulators asked for a presentation and time to review the model's assumptions and QA/QC.
Schultz said the model reduced unnecessary homeowner notifications in pilot runs and that the city is assembling its experts to walk state regulators through the methodology and the verification steps that tie model outputs back to inspection records.
