The emergence of big data has necessitated the development of high-performance computing engines capable of handling complex SQL workloads in distributed environments. SQL-on-Hadoop engines, such as Hive and Spark, have become critical in processing large-scale datasets within the Hadoop ecosystem. However, optimizing these engines for performance and resource efficiency remains a significant challenge due to the dynamic nature of big data environments. This paper presents a novel hybrid approach combining iterative refinement and machine learning-driven models for tuning configuration parameters in SQL-on-Hadoop systems. By systematically analyzing and adjusting resource allocation strategies, An optimization framework is proposed that enhances query processing times while minimizing computational costs. the approach introduces an automated tuning mechanism for adjusting system configurations in real-time, offering significant improvements over traditional manual tuning methods. Through extensive experimental evaluations on diverse workloads, the proposed model demonstrates superior scalability and efficiency, making it a valuable tool for cloud-based big data environments and next-generation data analytics applications.