Treffer: Artificial Intelligence-Based Performance Optimization of Oracle PL/SQL Queries ; Oracle PL/SQL Sorgularının Yapay Zekâ Tabanlı Performans Optimizasyonu

This study aims to develop an artificial intelligence–based optimization system to analyze and improve the performance of slow-running queries in Oracle PL/SQL and Forms-based applications. Performance data from Oracle queries were collected using SQL_TRACE and EXPLAIN PLAN and analyzed in a Python environment. A dataset was constructed through feature selection based on metrics such as execution time, logical reads, and I/O operations. Random Forest and XGBoost algorithms were applied to identify factors contributing to query slowness, with historical performance records used for model training and evaluation through standard performance metrics. The system was further refined and validated using real-world queries to enhance its recommendation capability. Results indicate substantial improvements: execution time reduced by 82.4%, consistent read rate by 84.8%, physical read rate by 90.9%, and total Oracle cost by 97%. In model comparison, XGBoost achieved superior classification accuracy with 96.1% accuracy and F1-score, while Random Forest provided faster prediction times. This research introduces a novel AI-driven system for diagnosing and optimizing Oracle PL/SQL performance issues, offering decision support for database administrators and contributing to improved query efficiency. ; Bu çalışma, Oracle PL/SQL ve Forms tabanlı uygulamalarda yavaş çalışan sorguları analiz ederek performansı artırmayı amaçlayan yapay zekâ destekli bir optimizasyon sistemi geliştirmeyi hedeflemektedir. Araştırmada, Oracle veri tabanında çalıştırılan sorguların performans verileri SQL_TRACE ve EXPLAIN PLAN kullanılarak toplanmış, Python ortamında analiz edilmiştir. Çalışma süresi, mantıksal okuma ve I/O işlemleri gibi metriklere dayalı özellik seçimiyle veri seti oluşturulmuş; sorgu yavaşlığının nedenlerini belirlemek için Random Forest ve XGBoost algoritmaları uygulanmıştır. Tarihsel performans kayıtlarıyla eğitilen modellerin doğruluğu çeşitli metriklerle değerlendirilmiş, sistem gerçek ortamdan alınan yeni sorgularla test ...