Total Organic Carbon (TOC) is a critical parameter for evaluating the enrichment of organic matter and the hydrocarbon generation potential of shale reservoirs. Conventional seismic TOC prediction techniques typically rely on high-accuracy density inversion. However, density inversion methods based on isotropic assumptions often yield substantial uncertainties in shale formations exhibiting strong Vertical Transverse Isotropy(VTI), thereby degrading the reliability of TOC estimation. Previous rock-physics studies have demonstrated that the VTI anisotropy parameters ε and σ exhibit strong correlations with TOC. In this study, we propose a novel TOC prediction methodology that integrates anisotropy-parameter seismic inversion with neural-network modeling. First, rock-physics analyses are conducted using laboratory measurements and wireline logs to elucidate the quantitative relationships between TOC and the anisotropy parameters. Second, an optimal three-term approximated reflection-coefficient equation for VTI media is selected to perform joint seismic inversion, and the anisotropy parameters ε and σ are subsequently derived through stepwise numerical computation. Finally, a Backpropagation (BP) neural network is constructed to establish a nonlinear mapping between TOC and the anisotropy parameters ε and σ, enabling seismic prediction of TOC. Synthetic tests indicate that the proposed approach maintains strong robustness under moderate noise levels. Application to field data shows that the inverted TOC results near wells exhibit good agreement with logging-derived TOC, thereby confirming the practicality and reliability of the method for quantitative characterization and spatial prediction of TOC in shale reservoirs.