Adaptive Neural Networks for predicting Research Excellence Framework results

Abstract

In many countries, the task of evaluating high-quality scientific research is performed by public agencies that produce an assessment of the activities carried out by universities and other High Education Providers (HEPs). This evaluation process is crucial as it constitutes the main basis of allocating national financial resources among HEPs.
In our contribution, we are using neural networks for predicting the next UK Research Excellence Framework exercise (REF) utcomes, i.e., a HEP’s rank and its quality-related research recurrent funding, and comparing our results with respect to other linear models in literature. We show that our approach lead to accurate results and success in predicting the outcomes of the UK’s research evaluation exercise. The benefits of predicting quality research are noteworthy and involve alike national funding agencies, educational institutions and researchers.