Hybrid learning in post-pandemic higher education systems: an analysis using SEM and DNN

The COVID-19 pandemic significantly impacted higher education systems, leading many institutions to adopt hybrid learning models. This research investigates the relevance of hybrid learning methods in the post-pandemic education context using structural equation modeling (SEM) and deep neural network (DNN) approaches. We tested the model at a public university in Indonesia that had implemented a hybrid learning system post-pandemic but reverted to a full offline system due to doubts about the benefits and student interest. Data were collected through questionnaires evaluating key factors, including social influence, perceived interactivity, perceived usefulness, ease of use, facility conditions, attitude, satisfaction, and user intention. SEM validated the conceptual model, confirming all eight hypotheses as statistically significant (p-value of ≤ 0.05) and achieving good model fit with a comparative fit index (CFI) of 0.901 and a root mean square error of approximation (RMSEA) of 0.069. Meanwhile, DNN achieved a high prediction accuracy of 82.40%, significantly outperforming logistic regression (baseline) models. The DNN demonstrated its ability to capture complex, nonlinear relationships and provide actionable insights into factors driving student interests. This research provides valuable empirical evidence to inform education policymakers, institutions, and stakeholders navigating the evolving landscape of post-pandemic learning environments.