<p>Accurate prediction of the spatial mechanism’s dynamic parameters in microgravity deployment simulations is crucial for identifying potential faults and ensuring precise gravitational compensation. Traditional engineering models are often inaccurate, primarily because of insufficient experimental data and incomplete understanding of physical phenomena, which impedes model bias reduction in information-poor scenarios. We present a novel hybrid approach aimed at improving the predictive accuracy of the dynamic behavior of spatial deployable mechanisms. The graph convolutional network-temporal convolutional network (GCN-TCN) model, a type of deep learning architecture, is utilized for its expertise in forecasting spatio-temporal data through multi-step predictions. Next, the adaptive bandwidth kernel density estimation technique is applied to estimate the probability density function of residuals from the testing set of the GCN-TCN, quantifying predictive uncertainty. The predictive information is further refined using Bayesian inference, integrating a priori knowledge from physics-based models with data from data-driven models to yield robust posterior predictions. The proposed methodology is validated and shown to be robust through rigorous numerical simulations and experimental validation, demonstrating its ability to provide accurate and reliable predictions for the deployment of spatial mechanisms.</p>