Akademik Veri Yönetim Sistemi
III. International Science and Innovation Congress , Tokat, Türkiye, 9 - 12 Haziran 2022, ss.1
As it is known, the medical sector is one of the areas where the most data is stored. With the development of technology in recent years, data stored in computers has reached gigantic proportions, especially in the field of medicine. Finding ways to extract meaningful information from these data in order to find solutions to patients' complaints provides great benefits for both patients and healthcare professionals. The leading of these methods is data mining techniques. Data mining is the process of reaching meaningful information by discovering certain relationships from data stacks through algorithmic structures based on statistical methods. Modeling based on the available data according to the concept of the problem encountered. Thus, it provides the opportunity for forward forecasting. It offers significant contribution to decision making mechanisms. In this context, the importance of diagnosing diseases in the field of medicine is increasing day by day. Specifically in pediatric surgery patients with hyoaspadios and chordee or patients who had circumcision operation frequently encountered. Hypospadias describes congenital absence of the urethral opening where it should be. Although it is 1 in 250 births in boys, it is the second most common congenital anomaly. Chordee is the congenital curvature of the penis in boys. Circumcision is the surgical removal of the tip of the skin covering the penis. In this study, supervised machine learning methods has been applied based on the data driven from results of the patients who were diagnosed with hypospadias, chordee or had circumcision operation. For this purpose, among 15533 patient data obtained from the pediatric surgery department, Laboratory results of 2871 patients were used. In the data preprocessing process data with inconsistent and incomplete abstract were identified and extracted. 70% of data is separated for training of machine learning models while 30% of the data used for testing. By comparing the results obtained from different methods in data mining the models with the best results were determined. With this work and with the help of models that determined by the laboratory results of diseases which have common complaints it is hoped that it will contribute to decision-support mechanisms.