Aspect-Based Sentiment Analysis Ulasan Pengguna Aplikasi Polri Presisi

Abror, Ahmad (2026) Aspect-Based Sentiment Analysis Ulasan Pengguna Aplikasi Polri Presisi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043221003_Undergraduate_Thesis.pdf - Updated Version Restricted to Repository staff only Download (15MB) | Request a copy

Abstract

Transformasi digital pada layanan publik mendorong pemanfaatan aplikasi mobile sebagai sarana peningkatan kualitas dan efisiensi pelayanan. Aplikasi Polri Presisi merupakan salah satu implementasi layanan kepolisian berbasis digital yang menghasilkan volume ulasan pengguna dalam jumlah besar. Penelitian ini bertujuan untuk menganalisis persepsi pengguna aplikasi Polri Presisi menggunakan pendekatan Aspect-Based Sentiment Analysis (ABSA) guna mengidentifikasi aspek utama layanan, mengklasifikasikan sentimen pada setiap aspek, serta menentukan prioritas perbaikan aplikasi. Data penelitian berupa 2460 ulasan pengguna aplikasi Polri Presisi versi 2.0.27 berbahasa Indonesia yang diperoleh dari Google Play Store. Identifikasi aspek dilakukan menggunakan metode Latent Dirichlet Allocation (LDA), sedangkan klasifikasi sentimen dilakukan menggunakan model IndoBERT. Selanjutnya, model Long Short-Term Memory (LSTM) digunakan untuk melakukan klasifikasi topik dan sentimen serta dievaluasi kinerjanya. Hasil penelitian menunjukkan bahwa LDA menghasilkan empat aspek utama, yaitu Pengurusan SKCK, Efisiensi Pendaftaran, Isu Teknis, serta Kepuasan Layanan. Analisis sentimen menunjukkan bahwa aspek Isu Teknis didominasi sentimen negatif dan memiliki nilai Net Reputation Score terendah sebesar -19,84%, sehingga menjadi prioritas utama perbaikan. Model LSTM menunjukkan performa yang baik dengan akurasi sebesar 85.07% pada klasifikasi topik dan akurasi sebesar 86.91% pada klasifikasi sentimen, meskipun masih mengalami keterbatasan dalam mengenali sentimen netral. Penelitian ini diharapkan dapat menjadi dasar evaluasi dan pengembangan berkelanjutan layanan kepolisian berbasis digital.
=================================================================================================================================
Digital transformation in public services has encouraged the adoption of mobile applications to improve service quality and accessibility. Polri Presisi is a digital-based police service application that generates a large volume of user reviews reflecting public perceptions of its performance. This study aims to analyze user opinions on the Polri Presisi application using an Aspect-Based Sentiment Analysis (ABSA) approach to identify key service aspects, classify sentiment for each aspect, and determine application improvement priorities. The dataset consists of 2,640 Indonesian-language user reviews of Polri Presisi version 2.0.27 collected from Google Play Store. Aspect identification was performed using Latent Dirichlet Allocation (LDA), while sentiment classification was conducted using the IndoBERT model. Furthermore, a Long Short-Term Memory (LSTM) model was implemented to classify both aspects and sentiments and to evaluate classification performance. The results show that LDA successfully identified four main aspects: Ease of SKCK Processing & Application Benefits, Online Registration Efficiency & Assistance, Technical Issues, Errors & Application Usage Constraints, and Service Satisfaction & Fast Access. Sentiment analysis reveals that the Technical Issues, Errors & Application Usage Constraints aspect is dominated by negative sentiment and has the lowest Net Reputation Score of -19.84%, indicating it as the highest priority for improvement. The LSTM model achieved a satisfactory performance with an accuracy of 85.07% for topic classification and an overall sentiment classification accuracy of 86.91%, although it demonstrated limitations in identifying neutral sentiment. This study provides data-driven insights to support continuous improvement of digital police services and enhance user satisfaction.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Aspect-Based Sentiment Analysis, IndoBERT, Latent Dirichlet Allocation (LDA), Long Short-Term Memory (LSTM), Net Reputation Score, POLRI PRESISI. Aspect-Based Sentiment Analysis, IndoBERT, Latent Dirichlet Allocation (LDA), Long Short-Term Memory (LSTM), Net Reputation Score, POLRI PRESISI
Subjects:	Q Science Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics Q Science > QA Mathematics > QA336 Artificial Intelligence
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Ahmad Abror
Date Deposited:	23 Jan 2026 02:04
Last Modified:	23 Jan 2026 02:04
URI:	http://repository.its.ac.id/id/eprint/130168

Actions (login required)

View Item