Abstract: Shear wave velocity serves as a key parameter for characterizing the elastic properties of rocks and their sensitivity to fluids, playing a significant role in seismic wave impedance inversion, stratigraphic identification, and rock physics modeling. However, the limitations in acquisition costs and technologies often lead to data gaps or insufficient accuracy for shear wave velocity. Hence, this study proposed an intelligent method for predicting shear wave velocity by integrating the two-dimensional convolutional neural network (2DCNN), a state-space model, and a cross-attention mechanism. The 2DCNN and the state-space model can capture local spatial features and longitudinal temporal dependencies, respectively, while the cross-attention mechanism enables the deep fusion of spatial and temporal features, thereby enhancing the sensitivity of the network to key geological information. Additionally, to improve model explainability, an explanatory analysis method based on Shapley additive explanations (SHAP) was employed to quantify the contributions of different log parameters to the shear wave velocity prediction. Experimental results demonstrate that the proposed method outperformed traditional individual neural networks in both prediction accuracy and generalization capability. Moreover, it effectively revealed the physical relationships between shear wave velocity and conventional log parameters. Therefore, the proposed method provides a novel approach and technical support for rapidly and reliably predicting shear wave velocity under complex reservoir conditions.