Gama, ShinduNata (2014) Clustering Portal Jurnal Internasional Untuk Rekomendasi Publikasi Berdasarkan Kualitas Cluster Menggunakan Kernel K-Means. Sarjana thesis, Universitas Brawijaya.
Abstract
Penelitian sebagai pendorong kemajuan teknologi, umumnya dilakukan pada lembaga-lembaga riset negara atau pada institusi. Penelitian dalam laboratorium tentu memerlukan referensi terkait dengan penelitian yang dilakukan. Umumnya, referensi tersebut berasal dari publikasi jurnal yang telah diterbitkan di seluruh dunia (internasional) dalam portal-portal jurnal. Semakin bagus kualitas jurnal-jurnal yang terdapat dalam suatu portal jurnal, maka semakin bagus pula kualitas portal jurnal tersebut. Oleh karena itu, peneliti memerlukan suatu sistem yang mampu memberikan rekomendasi portal jurnal internasional sehingga peneliti memperoleh referensi berupa jurnal yang berkualitas dan menghasilkan penelitian yang berkualitas pula. Salah satu teknik yang dapat dipakai adalah clustering. Metode dalam clustering sangat banyak tetapi yang paling sering digunakan adalah k-means. K-means sangat cepat dalam mengelompokkan data tetapi tidak mampu mengelompokkan data non-linierly separable sehingga digunakan metode kernel k-means dimana titik data akan dipetakan ke dimensi baru dengan fungsi pemetaan Φ sebelum dilakukan proses clustering. Dari hasil penelitian yang dilakukan, program mampu menghasilkan silhouette coefficient terbaik sebesar 0.853 dengan menggunakan 2 cluster, persentase sampel data yaitu 90%, serta menggunakan 3 fitur yaitu Total Docs. (3 years), Total Cites (3 years), dan Citable Docs. (3 years), dimana silhouette coefficient tertinggi adalah 1 yang berarti program mampu melakukan clustering terhadap data jurnal dengan baik.
English Abstract
Research as a driver of technological progress, is generally carried out at research institutions in the state or educational institutions such as universities and institutes. Researches in laboratory would require a reference related to the research conducted. Generally, the reference comes from journal publications that have been published worldwide (international) in the portals of the journal. The better the quality of the journals contained in a journal portal, then the better the quality of the journals portal.. Therefore, researchers require a system that is able to provide recommendations of international journal portal so that the researcher get high-quality journal and also produce high-quality research. One technique that can be used is clustering. Clustering method is very many but the most commonly used is the k-means clustering. K-means clustering very quickly in classifying the data but not able to classify the non-linierly separable data. To overcome the weakness of the method used kernel k-means clustering where data points are mapped into a new dimension with a higher mapping function Φ prior to the clustering process. From the research conducted, the program is able to produce the best silhouette coefficient of 0.853 by using two clusters, the percentage of minimum data for the sampling of data is 90%, and using three features, namely Total Docs. (3 years), Total Cites (3 years), and Citable Docs. (3 years) where the highest silhouette coefficient is 1, which means that the program is able to perform clustering of the journal data well.
Item Type: | Thesis (Sarjana) |
---|---|
Identification Number: | SKR/FTIK/2014/260/051500013 |
Subjects: | 000 Computer science, information and general works > 005 Computer programming, programs, data |
Divisions: | Fakultas Ilmu Komputer > Teknik Informatika |
Depositing User: | Budi Wahyono Wahyono |
Date Deposited: | 06 Jan 2015 09:21 |
Last Modified: | 20 Oct 2021 06:12 |
URI: | http://repository.ub.ac.id/id/eprint/146055 |
Preview |
Text
SHINDU_NATA_GAMA_-_105060800111093.pdf Download (11MB) | Preview |
Actions (login required)
![]() |
View Item |