PENERAPAN CONTENT-BASED FILTERING DAN NAÏVE BAYES CLASSIFIER PADA REKOMENDASI DOSEN PEMBIMBING TUGAS AKHIR

Sana, Bintang Jagad Syah (2025) PENERAPAN CONTENT-BASED FILTERING DAN NAÏVE BAYES CLASSIFIER PADA REKOMENDASI DOSEN PEMBIMBING TUGAS AKHIR. Undergraduate thesis, UPN "Veteran" Jawa Timur.

	Text (Cover) 21081010005_Cover.pdf Download (1MB)
	Text (Bab 1) 21081010005_ Bab 1 Pendahuluan.pdf Download (195kB)
	Text (Bab 2) 21081010005_Bab 2 Tinjauan Pustaka.pdf Restricted to Repository staff only until 15 September 2027. Download (348kB)
	Text (Bab 3) 21081010005_Bab 3 Metode Penelitian.pdf Restricted to Repository staff only until 15 September 2027. Download (903kB)
	Text (Bab 4) 21081010005_Bab 4 Hasil dan Pembahasan.pdf Restricted to Repository staff only until 15 September 2027. Download (1MB)
	Text (Bab 5) 21081010005_Bab 5.pdf Download (166kB)
	Text (Daftar Pustaka) 2108100005_Daftar Pustaka.pdf Download (152kB)
	Text (Lampiran) 21081010005_Lampiran.pdf Restricted to Repository staff only Download (753kB)

Abstract

The selection of thesis supervisors is a crucial factor that significantly influences students’ success in completing their undergraduate theses. Limited access to information regarding the alignment of lecturers’ expertise often leads to inaccurate supervisor assignments. This study proposes a supervisor recommendation system based on text mining by integrating Content-Based Filtering (CBF) and the Naïve Bayes Classifier (NBC). The dataset consists of lecturers’ scientific publications as proxies for their expertise and a collection of student thesis titles from 2022–2024 as test data. The preprocessing stage includes translation, case folding, tokenization, stopword removal, and stemming. CBF measures the similarity between thesis titles and publications using TF-IDF and cosine similarity, while NBC models the probability of topic relevance through Laplace smoothing. The scores from both methods are combined using a weighted sum approach to generate the final ranking of recommended supervisors. The evaluation results indicate that the hybrid CBF+NBC approach outperforms each method individually, showing a consistent improvement in accuracy across the years. Specifically, for the 2024 validation dataset, the system achieved 86% accuracy under the optimal weighting scheme, highlighting the effectiveness of combining content similarity and probabilistic modeling in mapping the compatibility between lecturers and students. This study provides a more objective, transparent, and balanced framework for supervisor selection, and it can be replicated in other academic programs with similar characteristics

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Puspaningrum, Eva Yulia	UNSPECIFIED	UNSPECIFIED
Thesis advisor	Haromainy, Muhammad Muharrom Al	UNSPECIFIED	UNSPECIFIED

Subjects:

Q Science > QA Mathematics > QA75 Electronic computers. Computer science

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Bintang Jagad Syah Sana

Date Deposited:

15 Sep 2025 08:23

Last Modified:

15 Sep 2025 08:23

URI:

https://repository.upnjatim.ac.id/id/eprint/43578

Actions (login required)

View Item