IDENTIFIKASI KEPAKARAN DOSEN BIDANG KOMPUTER DI UNIVERSITAS NEGERI JAKARTA MENGGUNAKAN LONG SHORT-TERM MEMORY DAN BIDIRECTIONAL LONG SHORT-TERM MEMORY

BAMBANG SETIAWAN MAULUDIN, . (2026) IDENTIFIKASI KEPAKARAN DOSEN BIDANG KOMPUTER DI UNIVERSITAS NEGERI JAKARTA MENGGUNAKAN LONG SHORT-TERM MEMORY DAN BIDIRECTIONAL LONG SHORT-TERM MEMORY. Sarjana thesis, UNIVERSITAS NEGERI JAKARTA.

	Text COVER.pdf Download (3MB)
	Text BAB 1.pdf Download (2MB)
	Text BAB 2.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB 3.pdf Restricted to Registered users only Download (4MB) | Request a copy
	Text BAB 4.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB 5.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text DAFTAR PUSTAKA.pdf Download (2MB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (4MB) | Request a copy

Abstract

Pemetaan kepakaran dosen di bidang komputer Universitas Negeri Jakarta (UNJ) saat ini belum terotomatisasi secara terstruktur. Penelitian ini bertujuan menganalisis dan membandingkan performa model deep learning Long Short-Term Memory (LSTM) dan Bidirectional LSTM (BiLSTM) untuk mengklasifikasikan publikasi ilmiah dosen berdasarkan standar ACM Computing Classification System (ACM CCS) 2020. Data yang digunakan berjumlah 149 dokumen publikasi terindeks Scopus dari dosen program studi Ilmu Komputer, PTIK, dan STI. Metode yang diterapkan meliputi pra-pemrosesan teks untuk mereduksi noise dan pemodelan topik Latent Dirichlet Allocation (LDA) guna mendukung proses pelabelan data secara hibrida. Fitur teks diekstraksi dan dibandingkan menggunakan representasi Word Embedding jenis Word2Vec dan FastText. Hasil evaluasi pada 129 dokumen valid (setelah penyaringan kategori NON-CC2020) menunjukkan bahwa FastText lebih superior dalam menangani variasi istilah teknis komputasi (Out-of-Vocabulary) dibandingkan Word2Vec. Kombinasi FastText dengan model LSTM searah menghasilkan performa paling optimal dengan tingkat akurasi 84,62% dan macro F1-score 67,73%. Sebaliknya, arsitektur BiLSTM justru menunjukkan penurunan akurasi (80,77%) akibat kejenuhan representasi (over-parameterization) pada data latih yang terbatas. Model terbaik ini secara efektif berhasil memetakan tren kepakaran dosen UNJ, yang saat ini didominasi oleh Data Science (22,15%), Information Technology (18,79%), dan Computer Science (17,45%). Hasil ini diharapkan dapat menjadi landasan penting untuk pengembangan institusi di masa depan. ***** The mapping of lecturers' expertise in the computer field at Universitas Negeri Jakarta (UNJ) was not structurally automated. This study aimed to analyze and compare the performance of Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) deep learning models in classifying lecturers' scientific publications based on the 2020 ACM Computing Classification System (ACM CCS) standard. The data used consisted of 149 Scopus-indexed publication documents from lecturers in the Computer Science, PTIK, and STI study programs. The applied methods included text preprocessing to reduce noise and Latent Dirichlet Allocation (LDA) topic modeling to support the hybrid data labeling process. Text features were extracted and compared using Word Embedding representations, specifically Word2Vec and FastText. The evaluation results on 129 valid documents (after filtering out the NON-CC2020 category) showed that FastText was superior in handling variations of computational technical terms (Out-of-Vocabulary) compared to Word2Vec. The combination of FastText with the unidirectional LSTM model yielded the most optimal performance with an accuracy rate of 84.62% and a macro F1-score of 67.73%. In contrast, the BiLSTM architecture exhibited a decrease in accuracy (80.77%) due to representational saturation (over-parameterization) on the limited training data. This best model effectively mapped the trends in UNJ lecturers' expertise, which was dominated by Data Science (22.15%), Information Technology (18.79%), and Computer Science (17.45%). These results were expected to serve as an important foundation for future institutional development.

Item Type:	Thesis (Sarjana)
Additional Information:	1). Dr. Widodo, M.Kom. ; 2). Murien Nugraheni, S.T. M.Cs.
Subjects:	Sains > Matematika > Ilmu Komputer
Divisions:	FT > S1 Pendidikan Teknik Informatika Komputer
Depositing User:	Users 34432 not found.
Date Deposited:	04 Aug 2026 06:28
Last Modified:	04 Aug 2026 06:28
URI:	http://repository.unj.ac.id/id/eprint/68188

Actions (login required)

View Item