Pemodelan Topik Pada Ulasan Pengguna Aplikasi M-Pajak Menggunakan Metode Bertopic

Megantara, Arindra (2026) Pemodelan Topik Pada Ulasan Pengguna Aplikasi M-Pajak Menggunakan Metode Bertopic. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043221115-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Implementasi transformasi digital pada layanan perpajakan melalui aplikasi M-Pajak merupakan bagian dari upaya Direktorat Jenderal Pajak (DJP) dalam meningkatkan efisiensi dan kemudahan layanan bagi wajib pajak. Namun, sejak penerapan dan integrasi sistem administrasi perpajakan berbasis CoreTax, aplikasi M-Pajak masih menghadapi berbagai kendala teknis yang tercermin dari banyaknya keluhan pengguna pada platform Google Play Store. Kondisi tersebut menunjukkan adanya kesenjangan antara tujuan digitalisasi layanan dan pengalaman pengguna di lapangan, sehingga diperlukan analisis berbasis data untuk memahami persepsi pengguna secara sistematis melalui pendekatan analisis sentimen dan pemodelan topik. Data yang digunakan berupa 2.311 ulasan pengguna yang dikumpulkan menggunakan teknik scraping. Analisis sentimen dilakukan dengan metode lexicon-based menggunakan Indonesian Sentiment Lexicon (InSet), kemudian dilanjutkan dengan klasifikasi menggunakan Support Vector Classifier. Selanjutnya, pemodelan topik diterapkan menggunakan metode BERTopic untuk mengidentifikasi tema utama pada ulasan sentimen positif dan negatif. Hasil penelitian menunjukkan bahwa ulasan pengguna didominasi oleh sentimen negatif sebesar 74,9%, sedangkan sentimen positif sebesar 25,1%, di mana sentimen negatif umumnya berkaitan dengan permasalahan teknis seperti kesulitan login, verifikasi akun, dan ketidakstabilan sistem, sementara sentimen positif mencerminkan apresiasi terhadap kemudahan layanan dan manfaat aplikasi. Model klasifikasi SVC menghasilkan akurasi sebesar 89% pada data pengujian. Pemodelan topik menggunakan BERTopic menunjukkan bahwa pada sentimen positif, topik dengan coherence score tertinggi 0,511 merepresentasikan kemudahan layanan dan manfaat aplikasi, sedangkan pada sentimen negatif, topik dengan coherence score tertinggi 0,598 merepresentasikan keluhan teknis dan ketidakstabilan aplikasi.
=======================================================================================================================================
The implementation of digital transformation in taxation services through the M-Pajak application is part of the Directorate General of Taxes (DGT) efforts to improve the efficiency and convenience of services for taxpayers. However, since the implementation and integration of the CoreTax-based taxation administration system, the M-Pajak application still faces various technical obstacles, as reflected in the large number of user complaints on the Google Play Store platform. This situation indicates a gap between the goal of service digitalization and the user experience in the field, requiring data-based analysis to systematically understand user perceptions through sentiment analysis and topic modeling approaches. The data used consisted of 2,311 user reviews collected using scraping techniques. Sentiment analysis was performed using a lexicon-based method with the Indonesian Sentiment Lexicon (InSet), followed by classification using Support Vector Classifier. Next, topic modeling was applied using the BERTopic method to identify the main themes in positive and negative sentiment reviews. The results showed that user reviews were dominated by negative sentiment at 74.9%, while positive sentiment was at 25.1%. Negative sentiment was generally related to technical issues such as login difficulties, account verification, and system instability, while positive sentiment reflected appreciation for the ease of service and benefits of the application. The SVC classification model produced an accuracy of 89% on the test data. Topic modeling using BERTopic showed that in positive sentiment, the topic with the highest coherence score of 0.511 represented the ease of service and benefits of the application, while in negative sentiment, the topic with the highest coherence score of 0.598 represented technical complaints and application instability.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis Sentimen, BERTopic, M-Pajak, Pemodelan Topik, Sentiment Analysis, Topic Modelling
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA75 Electronic computers. Computer science. EDP Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) T Technology > T Technology (General)
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Arindra Megantara
Date Deposited:	25 Jul 2026 04:34
Last Modified:	25 Jul 2026 04:34
URI:	http://repository.its.ac.id/id/eprint/132459

Actions (login required)

View Item