Bootstrapping: A Key To Reliable Service Line Material Predictions In Water Systems

The LCRR mandates lead service line has been mandated to regulate lead service lines in US water systems. Predictive models, like machine learning, aid in compliance, but it is crucial to have accurate information. Missing any lead in a service line (a false negative) or saying there's lead when there isn't (a false positive) can both have negative impacts. This is where bootstrapping can help. 

Bootstrapping evaluates model performance by repeatedly testing with varied data. Bootstrapping generates multiple datasets, allowing assessment of model performance under diverse conditions and estimating reliability for future unseen data. This technique enhances model transparency, aligning with responsible AI principles, and aids in model improvement. Read more about the benefits of embracing an iterative process.