As a critical component of rotating machinery, the accurate monitoring of bearing degradation plays a key role in ensuring mechanical reliability and reducing maintenance costs. However, existing prediction methods often rely on a single degradation indicator, neglecting the interrelatedness of multidimensional degradation feature data. To address this, we introduce a novel predictive model that integrates Bayesian Vector Autoregression (BVAR) with Gated Recurrent Unit (GRU) models to fully leverage the original multidimensional feature data and enhance forecasting accuracy. The BVAR is utilized to analyze the temporal correlations between multidimensional features and degradation indicators, while the GRU models the nonlinear evolution of each feature. An optimization strategy guided by training error integrates these analyses into a BVAR-GRU ensemble model. In the experimental section, the PRONOSTIA datasets are employed to validate our proposed method. The results demonstrate that compared to single-indicator prediction methods, our model shows significant advantages in terms of predictive error and accuracy. This study not only improves the precision of bearing degradation prediction but also provides a new technical strategy for the maintenance and operation of rotating machinery.