SELEKSI FITUR ALGORITMA GENETIKA DALAM KLASIFIKASI DATA REKAM MEDIS PCOS MENGGUNAKAN SVM

Abstract

A hormonal imbalance causes a woman with polycystic ovarian syndrome (PCOS) to have an ovum or egg that does not mature normally. It usually occurs during the reproductive period, but is often difficult to detect due to lack of awareness. Therefore, it is important to detect this condition early so that proper treatment or prevention can be done. One way to diagnose PCOS is through the use of medical data. In this study, 40 variables were used, including hormonal data, ultrasound results, and other medical information. The method used was Support Vector Machine (SVM), which is able to handle non-linear data with a kernel. To improve accuracy, features were selected using a genetic algorithm, which resulted in 19 significant variables. By applying the selected variables as input, the classification produced the best model with 94.26% accuracy, 87.57% sensitivity, and 97.52% specificity. Without the feature selection process, SVM classification only has an accuracy of 82.46%, sensitivity of 60.91%, and specificity of 97.25%. From the findings of this research, it can be seen that the genetic algorithm feature selection method can improve SVM classification performance. Keywords: Genetic Algorithm, Classification, PCOS, Feature Selection, SVM.