Hybrid Artificial Intelligence Models for Complex Decision Support: Architectural Syntheses and Operational Real-World Appraisals

Abstract

ABSTRACT
Modern high-stakes organizational environments—including autonomous industrial process controls, aerospace telemetry assessments, localized smart micro-grids, and decentralized clinical risk architectures—require highly stable, explainable, and fault-tolerant decision support mechanisms. Traditional connectionist models (such as deep neural networks) exhibit state-of-the-art predictive capabilities but lack explicit logical rules, suffer from extreme vulnerability to data noise, and present severe 'black-box' opacity barriers. Conversely, classical symbolic paradigms (such as rule-based expert systems and fuzzy logic controllers) provide absolute structural explainability but cannot learn patterns from high-dimensional unindexed data streams. To resolve this core technological bottleneck, the development of Hybrid Artificial Intelligence Systems has emerged as a critical computer science domain. This review paper provides a rigorous technical synthesis of hybrid models, analyzing Neuro-Symbolic integration, Adaptive Neuro-Fuzzy Inference Systems (ANFIS), and evolutionary optimization algorithms. The paper evaluates how combining data-driven pattern extraction with formalized logic networks scales system performance, reduces optimization latency, and guarantees baseline compliance boundaries under extreme data noise. The review isolates a distinct research gap: the lack of decentralized, open-source validation testbeds capable of continuously tracking real-time semantic drift across non-stationary multi-modal production networks. Finally, we lay out an operational framework to guide health, industrial, and public safety ministries in deploying verified hybrid AI systems safely within modern international risk management infrastructures.

KEYWORDS: Hybrid Artificial Intelligence; Neuro-Symbolic Integration; ANFIS; Decision Support Systems; Explainable AI; Algorithmic Robustness Verification.