Spatial indexing is critical for efficiently organizing and retrieving spatially referenced data in database management. This study introduces an improvement to the Waffle framework, a disk-based spatial index focusing on dynamic updates. The improvement incorporates a novel similarity measure that predicts which subtrees will update during maintenance, enhancing the adaptability to query loads. The Waffle framework, known for its adaptability in bulk loading, dynamic updates, and query processing, is further improved with this enhancement. The novel bulk loading method ensures optimal disk page utilization, improving query processing efficiency. Waffle prevents the creation of overlapping nodes during updates, preserving spatial data integrity and minimizing redundancies. The maintenance mechanism dynamically allocates resources based on the current query workload, ensuring robust performance. Experimental evaluation demonstrates the superior performance of Waffle, especially in KNN queries, over state-of-the-art spatial indexes. Despite its strengths, there remains a gap in the maintenance step, specifically in choosing the next subtree to update. To address this, the study proposes integrating AI models such as K-nearest neighbor (kNN), cosine algorithms (CA), and genetic algorithms (GA) to predict the next subtree to update. This integration aims to demonstrate significant improvements in query performance and resource efficiency, paving the way for self-tuning spatial indexing systems. The experiments showed that the enhanced E waffle outperforms the traditional waffle by reducing the required number of I/Os to 10% of its achievement.