Not Everyone Gets Sick the Same Way: Personalized Cybersickness Prediction via Bayesian Networks with Individual Susceptibility Modeling

Predicting the onset of cybersickness enables proactive mitigation strategies that minimize user discomfort, which is a key requirement for the widespread adoption of virtual reality (VR) in education, healthcare, training, and entertainment. Current prediction models leverage multimodal data such as visual scene parameters and physiological signals, yet often overlook individual differences that strongly influence cybersickness susceptibility, limiting their generalizability across diverse users. To address this limitation, we propose a personalized cybersickness prediction framework based on Bayesian Networks (BN) that incorporates individual factors, including age, gender, and prior VR experience. We introduce three personalization strategies: 1) direct integration of user attributes, 2) a two-stage pipeline that learns a latent susceptibility score, and 3) stratified models trained on homogeneous user groups. We validate our proposed method on two publicly available datasets (Maze and Simulation21). On the Maze dataset, our approach achieves a 14.27% relative improvement over the non-personalized Bayesian baseline and outperforms the state-of-the-art (SOTA) DeepTCN and LSTM models for cybersickness prediction by 44.59% and 48.07%, respectively. Our model also surpasses SOTA personalization-based deep learning methods, improving accuracy by 37.83% over the transfer learning strategy. We observe similar trends on the Simulation21 dataset, where our personalized framework yields a 13.40% accuracy gain over the non-personalized baseline, outperforms DeepTCN and LSTM by 9.16% and 10.82%, and improves accuracy by 3.89% over the transfer learning approach. Our results further demonstrate consistent improvements in F1 score, Cohen's κ, and Spearman correlation. Our method also maintains low inference latency, supporting interpretable and efficient probabilistic predictions in VR settings.