Explainable AI for Data-Driven Design of High-Dimensional Predictive Studies
Researchers have developed an Exploratory AI Recommender to aid in the design of high-dimensional predictive studies, particularly in healthcare. This framework uses flexible AI to identify complex data patterns and explainable AI techniques to generate recommendations for feature exclusion, non-linear terms, and feature interactions. When applied to predict patient falls, the system suggested excluding 23 features and including 221 interactions, leading to an improved C-index from 0.805 to 0.815. AI
IMPACT Enhances the interpretability and performance of predictive models in high-dimensional settings, potentially increasing clinical trust and adoption.