Çocuklardaki kompleks apandisiti anlama ve makine öğrenmesi algoritmalarıyla tahmin etme

This study aims to understand the complexity of appendicitis which is frequently seen in children and predict the presence of the disease based on several blood test values. The analysis is conducted among children that have abdominal pain and who are under 18 years old. Descriptive statistics demonstrated that there are differences and correlations between the red blood cell, thrombocyte, and c-reactive protein values which are potentially significant explanatory for the machine learning algorithms to capture, explain and predict the disease. Basic and most used linear, non-linear, and tree-based algorithms are used to predict both the existence of appendicitis and complex appendicitis in patients. A total of 71 models are created through the study and compared via classification model performance metrics. The best performing algorithm outperformed the Alvarado Score, pediatric appendicitis score (PAS), and RIPASA scores which are clinical scoring systems used to diagnose acute appendicitis in children. With the use of such a tool, unnecessary medications and surgeries can be avoided.