TY  - JOUR
AU  - Jr, Daniel Dasig
AU  - Dhar, Bablu Kumar
AU  - Pascua, Sonia M.
AU  - Austria, Milani
PY  - 2026
DA  - 2026/03/03
TI  - Federated Neuro-Symbolic Intelligence for Privacy-Preserving Data Analytics: A Next-Generation Framework for Real-Time and Industry Applications
JO  - PWU Journal of Research, Innovation, and Transformation
T2  - PWU Journal of Research, Innovation, and Transformation
JF  - PWU Journal of Research, Innovation, and Transformation
VL  - 1
IS  - 1
SP  - 28
EP  - 38
DO  - 10.62762/JRIT.2026.408887
UR  - https://www.icck.org/article/abs/JRIT.2026.408887
KW  - federated learning
KW  - neuro-symbolic artificial intelligence
KW  - privacy-preserving analytics
KW  - explainable AI
KW  - distributed intelligence
KW  - real-time systems
AB  - The growth of distributed, institutionally siloed data has created demand for analytics frameworks that ensure privacy, interpretability, and real-time decision support. While federated learning enables decentralized model training without raw data sharing, most existing approaches rely on opaque neural models and lack explicit reasoning capabilities. This paper proposes a Federated Neuro-Symbolic Intelligence (FNSI) framework that integrates federated learning with symbolic reasoning to address these limitations. The architecture combines local neural learning with symbolic constraint enforcement at the client level and a privacy-preserving coordination layer that aggregates encrypted updates and harmonizes distributed knowledge. Formal algorithms and theoretical analysis establish convergence, logical soundness, and data non-mobility. Experimental results under heterogeneous and non-IID data demonstrate that FNSI achieves accuracy comparable to centralized learning, improved robustness over standard federated averaging, and a substantial reduction in logical constraint violations. Visualization-based analyses further show how symbolic projection corrects infeasible neural outputs while preserving predictive fidelity. Overall, FNSI provides a balanced solution for trustworthy, cross-industry analytics in regulated environments.
SN  - 3071-4001
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Jr2026Federated,
  author = {Daniel Dasig Jr and Bablu Kumar Dhar and Sonia M. Pascua and Milani Austria},
  title = {Federated Neuro-Symbolic Intelligence for Privacy-Preserving Data Analytics: A Next-Generation Framework for Real-Time and Industry Applications},
  journal = {PWU Journal of Research, Innovation, and Transformation},
  year = {2026},
  volume = {1},
  number = {1},
  pages = {28-38},
  doi = {10.62762/JRIT.2026.408887},
  url = {https://www.icck.org/article/abs/JRIT.2026.408887},
  abstract = {The growth of distributed, institutionally siloed data has created demand for analytics frameworks that ensure privacy, interpretability, and real-time decision support. While federated learning enables decentralized model training without raw data sharing, most existing approaches rely on opaque neural models and lack explicit reasoning capabilities. This paper proposes a Federated Neuro-Symbolic Intelligence (FNSI) framework that integrates federated learning with symbolic reasoning to address these limitations. The architecture combines local neural learning with symbolic constraint enforcement at the client level and a privacy-preserving coordination layer that aggregates encrypted updates and harmonizes distributed knowledge. Formal algorithms and theoretical analysis establish convergence, logical soundness, and data non-mobility. Experimental results under heterogeneous and non-IID data demonstrate that FNSI achieves accuracy comparable to centralized learning, improved robustness over standard federated averaging, and a substantial reduction in logical constraint violations. Visualization-based analyses further show how symbolic projection corrects infeasible neural outputs while preserving predictive fidelity. Overall, FNSI provides a balanced solution for trustworthy, cross-industry analytics in regulated environments.},
  keywords = {federated learning, neuro-symbolic artificial intelligence, privacy-preserving analytics, explainable AI, distributed intelligence, real-time systems},
  issn = {3071-4001},
  publisher = {Institute of Central Computation and Knowledge}
}