Title: Symbolic Hybrid Knowledge-Based Systems: Integrating Knowledge-Based Reasoning and Machine Learning in Explainable AI
Authors: Amal Dwimah, Samy S. Abu-Naser
Volume: 9
Issue: 8
Pages: 80-84
Publication Date: 2025/08/28
Abstract:
Artificial Intelligence (AI) has achieved remarkable advances in recent decades, yet the fundamental trade-off between interpretability and predictive power remains unresolved. Symbolic Knowledge-Based Systems (KBS) provide transparent and logically consistent reasoning but often fail to adapt to large-scale and noisy datasets. Conversely, Deep Learning (DL) excels in handling high-dimensional, unstructured data but suffers from opacity, often being criticized as a "black-box" paradigm. To reconcile this dichotomy, neuro-symbolic systems have emerged, integrating symbolic logic with sub-symbolic machine learning. This paper presents a comprehensive review of neuro-symbolic hybrid approaches, highlighting their strengths, challenges, and applications. Furthermore, we propose and experimentally evaluate a hybrid neuro-symbolic model applied to lung cancer treatment prediction, integrating a rule-based reasoning engine with a Graph Neural Network (GNN) over medical knowledge graphs. Results demonstrate that the hybrid system achieves superior performance compared to standalone symbolic or deep learning models, attaining an accuracy of 91.3% with enhanced explainability, as validated by domain experts. The findings suggest that hybrid neuro-symbolic systems represent a viable pathway toward trustworthy, explainable, and accurate AI in critical domains such as healthcare.