Detecting and Optimizing Flawed Queries in Triplestore-Based Knowledge Systems Using Reinforcement Learning

This study uses Reinforcement Learning (RL) to predict flawed SPARQL Protocol and Resource Description Framework Query Language (SPARQL) queries and optimize them by suggesting structural modifications and tuning execution parameters in triplestore-based knowledge systems, verified on smart car parking and medical datasets. The RL agent is trained not only to detect poorly performing queries but also to recommend query rewrites that improve completeness and efficiency. Applied to a medical knowledge base focused on Fahr's disease, the system achieved an 89% detection accuracy and significantly reduced average query time and timeout rates, demonstrating the potential of Artificial Intelligence (AI) to enhance both the quality and speed of query processing in sensitive semantic web databases.