Implementasi Modified K-Nearest Neighbor (MKNN) untuk Deteksi Penyakit Anemia

Abstract

Anemia is a condition where the hemoglobin level in the human body drops below the normal threshold. It can cause several negative effects, such as delayed psychomotor development, a higher risk of infectious diseases, and in women, the possibility of premature birth. Therefore, early detection of anemia is essential to speed up treatment and recovery. One method that can support the diagnostic process is machine learning, particularly the Modified K-Nearest Neighbor (MKNN) algorithm. MKNN is an improved of standard KNN, incorporating additional steps such as validity calculation and weighted voting, which are not present in the original version. In this study, MKNN was applied to detect anemia and achieved an accuracy of 84% using a 75:25 train-test data split and k=5. The dataset was collected from Jemursari Hospital in Surabaya, consisting of 100 patient records. These records were used to evaluate the performance of the MKNN algorithm in anemia detection.