Perbandingan Kinerja Algoritma K-Medoids Dan K-Means Untuk Klasifikasi Penyakit Kanker Serviks

Abstract

K-Medoids and K-Means are unsupervised algorithms that use Cluster Distance Performance to group data. Cluster Distance Performance is a distance measurement method that can help an algorithm group objects based on the similarity of the variables. Several studies have shown that using the right Cluster Distance Performance can improve the performance of the algorithm in clustering. This study aims to compare the results of clustering cervical cancer datasets using the K-Medoids and K-Means Clustering methods. The cervical cancer dataset is 858 records and 36 attributes. The clusters produced by the two methods are 2 classes. K-Means is more effective in dealing with small data sizes. The K-Medoids algorithm model formed 361 data in the positive cluster and 473 data in the negative cluster, while the K-Means algorithm formed 308 data in the positive cluster and 526 data in the negative cluster. In the cervical cancer dataset using K-Medoids it showed a DBI result of -3.517, whereas using K-Means the evaluation results showed a result of -1.108. Thus, clustering using the K-Medoids Clustering method has better results compared to the K-Means Clustering method, because it produces a smaller DBI value of -1.108.

Keywords: Clustering; K-Means; K-Medoids; Cervical cancer.