Clustering Produktivitas Publikasi Ilmiah Dosen Perguruan Tinggi XYZ Menggunakan Metode Ensemble Berbasis Density Dan Mixed-Type Data

Arfianta, Puput (2025) Clustering Produktivitas Publikasi Ilmiah Dosen Perguruan Tinggi XYZ Menggunakan Metode Ensemble Berbasis Density Dan Mixed-Type Data. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003231007-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Produktivitas publikasi ilmiah merupakan salah satu output yang harus dicapai oleh perguruan tinggi menuju peringkat level dunia. Upaya perguruan tinggi XYZ dalam melakukan terobosan terkait peningkatan kuantitas dan kualitas publikasi belum optimal. Kendala yang masih ditemui yaitu belum meratanya keaktifan dosen dalam menuliskan publikasi, dimana masih adanya keterbatasan SDM riset yang masih tergantung pada sejumlah dosen tertentu yang menyebabkan beban publikasi ilmiah dan aktivitas penelitian terfokus pada kelompok dosen yang terbatas. Hal ini menimbulkan adanya kasus outlier pada data produktivitas publikasi ilmiah dosen. Variabel yang digunakan untuk analisis cluster terdiri dari variabel numerik dan variabel kategorik. Mengingat kompleksitas data campuran dan adanya kasus outlier, maka metrik jarak yang digunakan adalah Gower distance yang dikombinasikan dengan metode clustering DBSCAN dan HDBSCAN. Selain menggunakan metode clustering berbasis distance, yaitu DBSCAN-Gower Distance dan HDBSCAN-Gower Distance, penelitian ini juga membandingkan metode berbasis ensemble yaitu Cluster Ensemble Based Mixed Data Clustering (CEBMDC) DBSCAN-QROCK dan CEBMDC HDBSCAN-QROCK. Hasil analisis menunjukkan bahwa metode CEBMDC DBSCAN-QROCK menghasilkan kinerja terbaik dengan nilai Silhouette Coefficient tertinggi, membentuk tujuh klaster yang representative. Klaster-klaster ini mencerminkan segmentasi dosen dari kelompok dengan produktivitas publikasi ilmiah cenderung rendah hingga kelompok dosen unggul.
=================================================================================================================================
The productivity of scientific publications is one of the outputs that universities must achieve to reach world-class rankings. The efforts of XYZ University in making breakthroughs related to the improvement of the quantity and quality of publications have not been optimal. The challenge that is still encountered is the uneven activity of lecturers in writing publications, where there is still a limitation of research human resources that depend on a few specific lecturers, causing the burden of scientific publication and research activities to be concentrated on a limited group of lecturers. This results in the presence of outlier cases in the data on the scientific publication productivity of lecturers. The variables used for cluster analysis consist of numerical variables and categorical variables. Considering the complexity of mixed data and the presence of outlier cases, the distance metric used is Gower distance combined with the DBSCAN and HDBSCAN clustering methods. In addition to using distance-based clustering methods, namely DBSCAN-Gower Distance and HDBSCAN-Gower Distance, this study also compares ensemble-based methods, specifically Cluster Ensemble Based Mixed Data Clustering (CEBMDC) DBSCAN-QROCK and CEBMDC HDBSCAN-QROCK. The analysis results show that the CEBMDC DBSCAN-QROCK method yields the best performance with the highest Silhouette Coefficient value, forming seven representative clusters. These clusters reflect the segmentation of lecturers from groups with low scientific publication productivity to groups of outstanding lecturers.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Clustering, CEBMDC, DBSCAN, Gower Distance, HDBSCAN, Produktivitas Publikasi Ilmiah, QROCK, Clustering, CEBMDC, DBSCAN, Gower Distance, HDBSCAN, QROCK, Scientific Publication Productivity
Subjects:	Q Science > Q Science (General) Q Science > QA Mathematics Q Science > QA Mathematics > QA278.55 Cluster analysis
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Puput Arfianta
Date Deposited:	06 Aug 2025 02:31
Last Modified:	06 Aug 2025 02:31
URI:	http://repository.its.ac.id/id/eprint/127701

Actions (login required)

View Item