YAPAY ZEKA TABANLI ORACLE PL/SQL SORGU PERFORMANS OPTİMİZASYONU: RANDOM FOREST VE XGBOOST KARŞILAŞTIRMALI BİR YAKLAŞIM
ASES VI. INTERNATIONAL KAYSERI SCIENTIFIC STUDIES CONGRESS, Konya, Türkiye, 28 - 30 Kasım 2025, ss.243-244, (Özet Bildiri)
- Yayın Türü: Bildiri / Özet Bildiri
- Basıldığı Şehir: Konya
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.243-244
- Maltepe Üniversitesi Adresli: Evet
Özet
Developing an artificial intelligence–driven optimization system that analyzes slow-running queries in Oracle PL/SQL and Forms-based enterprise applications is aimed at enhancing overall system performance. Within the scope of this research, performance data of executed queries were collected using SQL_TRACE and EXPLAIN PLAN tools and subsequently analyzed in the Python environment. Through feature selection, a dataset was constructed based on key performance metrics such as execution time, consistent reads (CR), and physical reads (PR). To identify the underlying causes of query latency, Random Forest and XGBoost algorithms were applied. The models were trained using historical performance records and evaluated through multiple accuracy and classification metrics, achieving high predictive performance. The developed system was further improved through continuous testing with new queries obtained from real-world environments, enhancing its recommendation capability. According to the experimental results, the proposed approach achieved remarkable improvements, including an 82.4% reduction in execution time, an 84.8% improvement in consistent read rate, a 90.9% improvement in physical read rate, and a 97% reduction in total Oracle cost. In model comparison, XGBoost demonstrated superior classification accuracy with 96.1% accuracy and F1-score, whereas Random Forest provided faster prediction times. Overall, the proposed system represents an innovative decision-support framework capable of automatically detecting and optimizing performance bottlenecks in Oracle PL/SQL and Forms-based systems. It provides a practical roadmap for database administrators and enterprise software developers to monitor, analyze, and optimize query performance—ensuring greater efficiency, sustainability, and cost-effectiveness in enterprise data management.