The rapid growth of the electric vehicle industry has created an urgent demand for high-conductivityConductivity electrode materials. In this study, a deep learning model called the Data-Physics-Informed Neural NetworkNeural network (DPINNDPINN) was developed by integrating data and physical information into the loss function of a deep neural networkNeural network. This model enables an inverse prediction process—from performance metrics back to material features. Linear constraints were imposed through first-order equations derived from Pearson correlation coefficients (R). Initial analysis revealed no input features exhibited strong linear correlation with conductivityConductivity (|R|>0.7), indicating the necessity for domain-specific knowledge integration. To address this limitation, physics-informed constraints incorporating elemental composition rules, crystallographic structure principles, and thermodynamic stability criteria were systematically introduced. The enhanced DPINNDPINN framework demonstrated remarkable performance improvements: R2 reached 0.99, RMSE decreased by 112.32 S/m, and 14 novel orthorhombic-structured electrode materials were identified as promising candidates through inverse designInverse design. This approach provides a strategic framework for overcoming data scarcity challenges in electrode material development and holds broad applicability across diverse material discovery domains.