As there is an inherent conflict between the number of viable research projects and the availability of funding, decision-makers must prioritise research questions. One method to prioritise research is to identify the study that would provide the greatest economic benefit. The expected value of sample information (EVSI) is a Bayesian decision metric that can quantify the economic benefit of a study with a pre-determined sample size. However, it is unclear how to adapt EVSI to assess the benefit of a group sequential design (GSD). Thus, we first formally define EVSI for GSD. We then employ machine learning techniques to develop novel EVSI estimation methods to efficiently compute EVSI for GSD. Their performances are compared with the conventional estimation approach, which is accurate but infeasibly complex in practice, using a real-world case study. The results show that the proposed methods can cut the computational time from two days to several seconds without a loss of accuracy.