Pengembangan Modul Rekomendasi E-Course berbasis Knowledge Rules pada Platform Tuladha

Syahputra, Muhammad Fahmi (2026) Pengembangan Modul Rekomendasi E-Course berbasis Knowledge Rules pada Platform Tuladha. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221302-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Pengasuhan anak pada masa keemasan (golden age) usia 0-5 tahun merupakan fondasi krusial bagi perkembangan karakter. Penelitian ini bertujuan untuk mengembangkan materi edukasi pengasuhan (e-course) pada platform Tuladha, namun navigasi kategori yang statis pada platform Tuladha berisiko memicu kelebihan beban kognitif (cognitive overload) bagi pengguna baru dalam mencari materi yang tepat. Penelitian ini mengembangkan modul rekomendasi guna memberikan intervensi edukasi terpersonalisasi untuk menangani ketiadaan riwayat interaksi awal (cold-start). Metode usulan ini menggabungkan pendekatan berbasis aturan dengan pencarian semantik. Parameter usia anak dan respons kuesioner Parenting Styles and Dimensions Questionnaire (PSDQ) diekstraksi menjadi kueri teks terstruktur. Kueri tersebut kemudian ditransformasikan ke ruang vektor menggunakan model Transformer untuk dihitung kemiripan kosinusnya terhadap pangkalan data materi e-course. Evaluasi sistem dilakukan melalui penilaian kepakaran (expert validation) terhadap 36 skenario kasus persona yang terdistribusi seimbang. Pengujian membandingkan kinerja arsitektur SBERT, Multilingual-E5, dan DistilBERT-Indonesian menggunakan metrik Hit@3, Overall Category Match Rate, dan Cohen's Kappa. Hasil eksperimen menunjukkan DistilBERT-Indonesian menghasilkan relevansi dokumen terbaik dengan Hit@3 sebesar 91,67% dan Overall Category Match Rate 87,96%. Sementara itu, Multilingual-E5 memperoleh tingkat kesepakatan tertinggi dengan nilai Cohen's Kappa 0,3438. Temuan ini mengindikasikan bahwa fusi antara logika ekstraksi kepakaran dan penelusuran semantik dapat menghasilkan rekomendasi preskriptif terarah dalam mengatasi permasalahan cold-start pada platform edukasi.
========================================================================================================================================
Parenting during the golden age of 0-5 years is a crucial foundation for character development. This study aims to develop parenting educational materials (e-course) within the Tuladha platform. However, static category navigation on the Tuladha educational platform risks triggering cognitive overload for new users seeking appropriate materials. Therefore, this research develops a hybrid recommendation module to provide personalized educational interventions while addressing the lack of initial user interaction history (the cold-start problem). The proposed method combines a rule-based approach with semantic search. Child age parameters and Parenting Styles and Dimensions Questionnaire (PSDQ) responses are extracted into structured text queries. These queries are then transformed into a vector space using a Transformer model to compute their cosine similarity against the e-course material database. System evaluation was conducted through expert judgment on 36 evenly distributed persona case scenarios. The test compared the performance of SBERT, Multilingual-E5, and DistilBERT-Indonesian architectures using Hit@3, Overall Category Match Rate, and Cohen's Kappa metrics. Experimental results show that DistilBERT-Indonesian produced the best document relevance with a Hit@3 of 91.67% and an Overall Category Match Rate of 87.96%. Meanwhile, Multilingual-E5 achieved the highest agreement level with a Cohen's Kappa of 0.3438. These findings indicate that fusing expert logic extraction and semantic retrieval can produce targeted prescriptive recommendations to overcome the cold-start problem on educational platforms.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	cold-start, golden age, parenting, PSDQ, semantic similarity, transformers, sistem rekomendasi, recommender system
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA76.76.E95 Expert systems Q Science > QA Mathematics > QA76.9.I58 Recommender systems (Information filtering)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Fahmi Syahputra
Date Deposited:	28 Jul 2026 01:31
Last Modified:	28 Jul 2026 01:31
URI:	http://repository.its.ac.id/id/eprint/138225

Actions (login required)

View Item