Comparison of K-Means and K-Medoids Algorithms in Text Mining based on Davies Bouldin Index Testing for Classification of Student’s Thesis

Abstract

The thesis is one of the scientific works based on the conclusions of field research or observations compiled and developed by students as well as research carried out according to the topic containing the study program which is carried out as a final project compiled in the last stage of formal study. A large number of theses, of course, will be sought in looking for categories of thesis topics, or the titles raised have different relevance. However, the student thesis can be by topics that are almost relevant to other topics so that it can make it easier to find topics that are relevant to the group. One of the uses of techniques in machine learning is to find text processing (Text Mining). In-text mining, there is a method that can be used, namely the Clustering method. Clustering processing techniques can group objects into the number of clusters formed. In addition, there are several methods used in clustering processing. This study aims to compare 2 cluster algorithms, namely the K-Means and K-Medoids algorithms to obtain an appropriate evaluation in the case of thesis grouping so that the relevant topics in the formed groups have better accuracy. The evaluation stage used is the Davies Bouldin Index (DBI) evaluation which is one of the testing techniques on the cluster. In addition, another indicator for comparison is the computation time of the two algorithms. According to the DBI value test carried out on algorithm 2, the K-Medoids algorithm is superior to K-Means, where the average DBI value produced by K-Medoids is 1,56 while K-Means is 2,79. In addition, the computational time required in classifying documents is also a reference. In testing the computational time required to group 50 documents, K-Means is superior to K-Medoids. K-Means has an average computation time for grouping documents, which is 1 second, while K-Medoids provide a computation time of 26,7778 seconds.

Rekomendasi Restock Barang di Toko Pojok UMKM Menggunakan Algoritma K-Means Clustering

“…DBI ialah pengujian untuk mengevaluasi, dan mengukur keakuratan sebuah cluster pada suatu algoritma Clustering [16], [17]. Pengujian menggunakan DBI diartikan sebagai rasio rata-rata jarak dalam, dari sebuah cluster satu dengan cluster lainnya dilihat dari jarak cluster terdekat antar tetangganya [18], [19].…”

“…Pengujian menggunakan DBI diartikan sebagai rasio rata-rata jarak dalam, dari sebuah cluster satu dengan cluster lainnya dilihat dari jarak cluster terdekat antar tetangganya [18], [19]. Tujuan dari metode pengujian DBI adalah untuk mengoptimalkan jarak antara satu cluster dengan cluster lainnya dan mencari nilai yang memperkecil jarak antara data dokumen yang terdapat dalam cluster yang sama [16], [20], [21].…”

Analisis Pengelompokkan Dataset Pemilu 2014 Dan 2019 DPR Ri Di Kota Cirebon Menggunakan Algoritma K-Means Clustering

“…Hasil penelitian menunjukkan bahwa penelitian ini tidak sepenuhnya fokus pada kinerja nilai parameter lain yang mungkin mempengaruhi kinerja algoritma K-Means. Namun, terdapat potensi untuk penelitian lebih lanjut dalam konteks metode Clustering yang memungkinkan pengguna menentukan nilai parameter dengan lebih jelas untuk meningkatkan kualitas DBI [11].…”

Implementasi Data Mining Menggunakan Algoritma K- Means Clustering Dalam Analisis Penjualan Produk

“…Pengujian DBI dilakukan dengan mengukur kekuatan atau keakuratan jumlah K (cluster) yang dibentuk menggunakan algoritma K-Means dan K-Medoids. [8]…”