Hybrid AI Architectures for Complex Decision Support Systems

Abstract

Hybrid AI architectures are emerging as a powerful solution for complex decision support systems (DSS) that must handle uncertainty, heterogeneous data, and large-scale integration. Traditional AI methods are limited in addressing real-world multidimensional challenges. This paper analyzes hybrid AI systems that combine symbolic approaches (rule-based reasoning, interpretability) with sub-symbolic methods (machine learning, neural networks) to improve flexibility and robustness. A modular framework is proposed, consisting of data preprocessing, knowledge representation, inference, and learning components, enabling both offline training and real-time decision-making. The study highlights key challenges such as scalability, knowledge integration, and computational efficiency. Experimental results demonstrate that hybrid models outperform standalone AI techniques in accuracy, precision, recall, and efficiency, especially in dynamic and uncertain environments. The paper concludes by suggesting future directions, including explainable AI and scalable distributed architectures.