An Adaptive Dual-Score Coevolutionary Algorithm for Personalized Exercise Group Assembly

Personalized Exercise Group Assembly (PEGA) is an essential strategy in modern education that should be considered. Usually, a PEGA task is a constrained large-scale sparse multi-objective problem (LSMOP). However, although many cognitive diagnostic methods have successfully revealed students’ abilities and some studies have utilized diagnostic results for exercise recommendation, they perform poorly when faced with the LSMOP, making it challenging to balance performance and efficiency. In this study, we develop an adaptive dual-score coevolutionary algorithm (ADSCA) to solve the problem. The adaptive and dual-score mechanisms of ADSCA have enhanced its performance in performing PEGA tasks. Experiments conducted on two datasets demonstrate that our proposed algorithm significantly outperforms the previous algorithm and exhibits excellent performance compared to mainstream exercise recommendation algorithms.