Although vertigo is a recognized complication of SARS-CoV-2 infection, its correlation with altered brain function remains unexplored. In this study, we utilized resting-state fMRI to investigate the cerebellum in eight patients experiencing persistent central vertigo following COVID-19, comparing them to 16 healthy controls. We conducted a whole-brain voxel-wise analysis, followed by a cerebellar map-based analysis, revealing significant between-group differences in Amplitude of Low-Frequency Fluctuations (ALFF), Regional Homogeneity (ReHo), and functional connectivity (FC), all localized to the cerebellum. We observed that the vertigo group, among other changes, showed reduced synchronization of neural activity in the flocculonodular lobe, a cerebellar region crucial for coordinating balance and eye movements. Next, we applied a machine learning algorithm to determine whether cerebellar changes related to SARS-CoV-2 exhibited distinct patterns, enabling effective classification of study participants as either vertigo-affected or healthy. The algorithm demonstrated outstanding discriminatory power with an average Area Under the Curve score of 0.901 (95% CI: 0.894 - 0.908). Consequently, we employed the SHapley Additive exPlanations approach to explain which factors were mostly responsible for our algorithm’s performance. Interestingly, we found that, rather than changes to the flocculonodular lobe, heightened neural activity in cerebellar regions associated with motor imagery, self-projection, and spatial simulation was of gratest impotance for our model. Nevertheless, incorporating data from additional cerebellar regions, analyzed using various other methods such as fractional ALFF, ReHo, and FC, further enhanced its capability, highlighting the complexity and multifaceted nauture of SARS-CoV-2-related changes.