PREEMPT
Project on Ensemble Model Complexity
How do different fields build and talk about “ensemble” models? Can we learn or help anything by comparing and contrasting parallel approaches to mixing data, model inputs, models, model outcomes, and model interpretations?
Is there a complexity in factors captured/weighted by ensemble predictive models that suddenly allows for drastically improved predictive power over models with fewer factors or different factorial weighting?
A growing applied forecasting literature indicates that model ensembles tend to outperform individual models. This is certainly the case in epidemic forecasting studies across different disease systems such as Ebola, dengue, and COVID-19.
However, there is a need to systematically investigate how different mechanisms, statistical, numerical, and data-related factors contribute to forecasting performance.
Check out the preprint of our first paper here!
Team Members
(alphabetical)
Publications & Products
Our first preprint!