US FDA’s Machine Learning Analysis Suggests Where To Look For Manufacturing Site Quality Problems

XGBoost model sees site complexity contributes to quality risk, but misses device-related issues, says third FDA pharmaceutical quality report.

Watching the machine learn
watching the machine learn • Source: Alamy

Using a machine learning regression model on a decade of manufacturing site inspection data, the US Food and Drug Administration’s Office of Pharmaceutical Quality has identified an array of underlying quality factors, some of them obvious to most observers, others quite obscure, that the agency may use in targeting future surveillance activities.

After chewing on the FDA’s data, the open-source model confirmed, for example, the widely held belief that manufacturing sites in...

More from Manufacturing

More from Compliance