Implementasi Sistem Rekomendasi Berbasis Web Menggunakan Content-Based Filtering dan Multiattribute Network Analysis

Rhakan, Muhammad (2025) Implementasi Sistem Rekomendasi Berbasis Web Menggunakan Content-Based Filtering dan Multiattribute Network Analysis. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211166-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Sistem rekomendasi memiliki peran penting dalam meningkatkan relevansi kursus pada platform pembelajaran elektronik seperti Udemy. Tugas akhir ini mengembangkan sistem rekomendasi berbasis konten (content-based filtering) dengan pendekatan analisis jaringan multiatribut yang sepenuhnya bergantung pada metadata kursus. Fitur teks, numerik, dan kategorikal diekstraksi dari data kursus, lalu dihitung tingkat keserupaannya menggunakan cosine similarity, euclidean distance, dan dice similarity. Nilai-nilai tersebut digabungkan melalui agregasi berbobot untuk membentuk matriks keserupaan, yang kemudian digunakan sebagai dasar dalam pembentukan jaringan berbobot dan tidak berarah. Struktur komunitas dalam jaringan dibentuk menggunakan algoritma Leiden, menghasilkan 262 komunitas. Evaluasi menunjukkan kualitas struktur yang sangat baik dengan nilai modularitas 0,992, cakupan dan performa 0,996, serta konduktansi 0,004. Rekomendasi diberikan berdasarkan nilai ego-focused centrality (CEF) tertinggi terhadap kursus target yang dipilih pengguna, dan ditampilkan dalam antarmuka web sebagai sepuluh kursus yang relevan secara kontekstual.
===================================================================================================================================
Recommender systems play a crucial role in enhancing content relevance on e-learning platforms such as Udemy. This thesis develops a content-based filtering approach using multiattribute network analysis, relying solely on metadata from course content. Textual, numerical, and categorical features are extracted and their similarities are computed using cosine similarity, euclidean distance, and dice similarity, respectively. These similarity values are then combined through a weighted aggregation scheme to form a final similarity matrix, which serves as the basis for constructing a weighted, undirected network. Community structures within the network are identified using the Leiden algorithm, resulting in 267 communities. Evaluation shows excellent structural quality, with modularity of 0.992, coverage and performance of 0.996, and conductance of 0.004. Recommendations are generated based on the ego-focused centrality (CEF) of each candidate node in relation to a user-selected target node, and displayed through a simple web interface as ten contextually relevant course suggestions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Pembelajaran Elektronik, Pendekatan Berbasis Konten, Sistem Rekomendasi, Analisis Jaringan, Analisis Deteksi Komunitas E-Learning, Content-Based Filtering, Recommendation System, Network Analysis, Community-Detection Analysis
Subjects:	Q Science > QA Mathematics > QA76.9.I58 Recommender systems (Information filtering)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Rhakan
Date Deposited:	28 Jul 2025 04:00
Last Modified:	28 Jul 2025 04:00
URI:	http://repository.its.ac.id/id/eprint/122532

Actions (login required)

View Item