Management Control Systems in Aviation: A Case Study of Turkish Airlines

This study examines the generalization performance of artificial intelligence-based regression models used in medical diagnosis and risk prediction. Three real-world datasets were utilized: Autism Screening Adult and Acute Inflammations from UCI Machine Learning Repository, and a Heart Disease Diagnosis dataset from OpenML. Regression models were preferred over complex classification algorithms and deep learning techniques due to their interpretable and transparent structure, which offers significant advantages in medical decision-making processes. Both simple and complex regression approaches were compared, including linear models such as linear regression, ridge, and lasso, as well as non-linear models such as SVR, decision tree-based regressions, random forest, and gradient boosting trees. Model performance was evaluated using generalization metrics, and factors affecting generalization success were identified. Findings indicate that regression-based decision support systems can produce meaningful predictions with medical data while maintaining transparency and accountability in clinical applications. Keywords: Artificial intelligence, regression analysis, decision support system, generalization, medical diagnosis, risk prediction, machine learning, supervised learning