OPTIMALISASI CHATBOT HELPDESK LAYANAN TIK DINAS KOMINFO BONE BOLANGO BERBASIS SEQ2SEQ ENCODER BILSTM DAN RETRIEVAL

Lakoro, Retno Reynaldi (2026) OPTIMALISASI CHATBOT HELPDESK LAYANAN TIK DINAS KOMINFO BONE BOLANGO BERBASIS SEQ2SEQ ENCODER BILSTM DAN RETRIEVAL. Sarjana thesis, Universitas Muhammadiyah Gorontalo.

[img] Text (Abstrak Retno Reynaldi Lakoro)
ABSTRACT.pdf

Download (512kB)
[img] Text (Bab Retno Reynaldi Lakoro)
BAB I.pdf

Download (160kB)
[img] Text (Daftar Pustaka Retno Reynaldi Lakoro)
DAFTAR PUSTAKA.pdf

Download (166kB)
[img] Text (Skripsi Retno Reynaldi Lakoro)
SKRIPSI - RETNO REYNALDI LAKORO.pdf
Restricted to Repository staff only

Download (7MB)

Abstract

The Information and Communication Technology (ICT) helpdesk service at the Bone Bolango Regency Communication and Informatics Office still faces challenges such as limited service hours, dependence on staff availability, and inconsistent responses to repetitive questions. These conditions result in inconsistent response times, potentially reducing service effectiveness, and making it difficult to meet ICT information needs quickly and consistently within the local government environment. This study aims to evaluate and optimize the performance of an ICT helpdesk chatbot in understanding the context of user questions, classifying service categories, and selecting relevant and consistent answers within a closed public service domain. The approach used was a quantitative method with an experimental design. The chatbot model was developed using a hybrid Sequence-to-Sequence (Seq2Seq) architecture with a Bidirectional Long Short-Term Memory (BiLSTM) encoder and an attention mechanism, along with a response-based retrieval approach for answer selection. The dataset consists of 7,192 question-answer pairs manually compiled based on real-world ICT service cases and grouped into four service categories. The analysis technique was performed through model performance evaluation using accuracy, precision, recall, F1-score, top-3 accuracy, and confusion matrix analysis to assess classification stability and accuracy. The results show that the model is capable of achieving classification accuracy approaching 99% with a high F1-score and stable top-3 accuracy on the test data, thus proving that the BiLSTM encoder-centric approach with an effective retrieval mechanism produces relevant, consistent, and controlled responses for application in a closed-domain public service helpdesk chatbot.

Item Type: Thesis (Sarjana)
Contributors:
ContributionContributorsNIDN/NBMEmail
Thesis advisor-, Syahrial0918088101UNSPECIFIED
Thesis advisorPranata, Widya Eka1626119801UNSPECIFIED
Additional Information: Layanan helpdesk Teknologi Informasi dan Komunikasi (TIK) di Dinas Komunikasi dan Informatika Kabupaten Bone Bolango masih menghadapi permasalahan berupa keterbatasan jam layanan, ketergantungan pada ketersediaan petugas, serta inkonsistensi jawaban terhadap pertanyaan yang bersifat berulang. Kondisi tersebut menyebabkan waktu respons yang tidak seragam, berpotensi menurunkan efektivitas pelayanan, serta menyulitkan pemenuhan kebutuhan informasi TIK secara cepat dan konsisten pada lingkungan pemerintahan daerah. Penelitian ini bertujuan untuk mengevaluasi dan mengoptimalkan kinerja chatbot helpdesk TIK dalam memahami konteks pertanyaan pengguna, mengklasifikasikan kategori layanan, serta memilih jawaban yang relevan dan konsisten pada domain layanan publik tertutup. Pendekatan yang digunakan adalah metode kuantitatif dengan desain eksperimen, di mana model chatbot dikembangkan menggunakan arsitektur hibrid Sequence-to-Sequence (Seq2Seq) dengan encoder Bidirectional Long Short-Term Memory (BiLSTM) dan mekanisme attention, serta pendekatan retrieval-based response dalam pemilihan jawaban. Dataset terdiri atas 7.192 pasangan pertanyaan–jawaban yang disusun secara manual berdasarkan kasus nyata layanan TIK dan dikelompokkan ke dalam empat kategori layanan, sedangkan teknik analisis dilakukan melalui evaluasi kinerja model menggunakan metrik akurasi, precision, recall, F1-score, top-3 accuracy, serta analisis confusion matrix untuk menilai kestabilan dan ketepatan klasifikasi. Hasil penelitian menunjukkan bahwa model mampu mencapai akurasi klasifikasi mendekati 99% dengan nilai F1-score yang tinggi dan top-3 accuracy yang stabil pada data uji, sehingga membuktikan bahwa pendekatan encoder-centric BiLSTM dengan mekanisme retrieval efektif menghasilkan respons yang relevan, konsisten, dan terkontrol untuk diterapkan pada chatbot helpdesk layanan publik berbasis domain tertutup.
Uncontrolled Keywords: BiLSTM, chatbot, helpdesk TIK, retrieval, Seq2Seq
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QA Mathematics > QA76 Computer software
Divisions: Fakultas Sains dan Ilmu Komputer > Program Studi Ilmu Komputer
Depositing User: Retno Reynaldi Lakoro
Date Deposited: 09 Jul 2026 07:01
Last Modified: 09 Jul 2026 07:01
URI: http://repository.umgo.ac.id/id/eprint/6730

Actions (login required)

View Item View Item