<![CDATA[Skripsi merupakan karya ilmiah mahasiswa sarjana berdasarkan penelitian dalam bidang tertentu. Di Fakultas Sains Institut Teknologi Sumatera (Itera), jumlah skripsi yang terus meningkat belum didukung system pengelompokan topik yang sistematis. Penelitian ini menerapkan metode Agglomerative Hierarchical Clustering (AHC) untuk mengelompokkan skripsi berdasarkan kemiripan topik. Data yang digunakan berupa 575 judul skripsi dari sembilan program studi di Fakultas Sains ITERA tahun 2024. Tahapan penelitian meliputi preprocessing teks, perhitungan bobot dengan metode TF-IDF, dan reduksi dimensi menggunakan Principal Component Analysis (PCA). Pengelompokan dilakukan dengan tiga metode linkage, yaitu average, single, dan complete, serta pengukuran kemiripan menggunakan cosine distance. Hasil menunjukkan bahwa metode average linkage memberikan hasil terbaik dengan nilai silhouette coefficient sebesar 0.3091 pada titik potong 0.86. Topik tiap cluster ditentukan dari lima kata kunci dominan berdasarkan nilai TF-IDF tertinggi sebagai label representatif. Penelitian ini diharapkan menjadi langkah awal dalam pengembangan sistem pengelompokan topik skripsi yang lebih terstruktur dan informatif. Analysis of Student Thesis Clustering in the Faculty of Science, Sumatra Institute of Technology, Using the Agglomerative Hierarchical Clustering Method Type your abstract here (10 pt). spasi 1Abstracts are written in two languages, namely Indonesian and English, typed in 1 paragraph 1 space of 150-250 words, containing research points, such as objectives, methods and research results. An undergraduate thesis is a scientific work by students based on research in a specific field. At the Faculty of Science, Institut Teknologi Sumatera (ITERA), the increasing number of theses has not yet been supported by a systematic topic grouping system. This study applies the Agglomerative Hierarchical Clustering (AHC) method to group theses based on topic similarity. The data used consists of 575 thesis titles from nine study programs at the Faculty of Science ITERA in 2024. The research stages include text preprocessing, term weighting using the TF-IDF method, and dimensionality reduction using Principal Component Analysis (PCA). Clustering was performed using three linkage methods: average, single, and complete, with similarity measurement using cosine distance. The results show that the average linkage method provided the best result with a silhouette coefficient value of 0.3091 at a cutting point of 0.86. The topic of each cluster was determined based on five dominant keywords with the highest TF-IDF values as representative labels. This study is expected to serve as an initial step in developing a more structured and informative thesis topic grouping system]]>
Copyrights © 2026
