Pengembangan Dashboard Analitik Pemodelan Topik Menggunakan LDA Dan Klasifikasi Sentimen Menggunakan Bi-LSTM Pada Ulasan Pengguna Aplikasi FlyGaruda

Subagja, Khurotaayun Pesona (2026) Pengembangan Dashboard Analitik Pemodelan Topik Menggunakan LDA Dan Klasifikasi Sentimen Menggunakan Bi-LSTM Pada Ulasan Pengguna Aplikasi FlyGaruda. Other thesis, Institut teknologi Sepuluh Nopember.

Text 2043221089_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Sebagai negara kepulauan yang luas, kebutuhan masyarakat Indonesia terhadap transportasi udara terus meningkat setiap tahunnya. Kondisi ini menjadikan industri penerbangan memiliki peran penting dalam mendukung mobilitas dan pertumbuhan ekonomi nasional. Garuda Indonesia, sebagai maskapai nasional dan anggota aliansi global SkyTeam, terus berupaya meningkatkan kualitas layanan digital melalui aplikasi FlyGaruda. Meskipun telah diunduh lebih dari satu juta kali di Google Play Store, aplikasi ini masih menerima berbagai keluhan dari pengguna. Oleh karena itu, penelitian ini bertujuan menganalisis persepsi pengguna terhadap aplikasi FlyGaruda melalui analisis sentimen dan pemodelan topik. Analisis sentimen dilakukan menggunakan pendekatan leksikon InSet dan metode Bidirectional Long Short-Term Memory (Bi-LSTM), sedangkan pemodelan topik menggunakan Latent Dirichlet Allocation (LDA). Hasil pemodelan topik menunjukkan dua topik optimal dengan nilai koherensi 0,4802, yaitu Penggunaan Layanan Maskapai Garuda Indonesia dan Fitur Login dan Akses Akun pada FlyGaruda. Pelabelan sentimen terhadap 1.685 ulasan menghasilkan 296 ulasan (17,6%) positif, 1.326 ulasan (78,7%) negatif, dan 63 ulasan (3,7%) netral. Sebanyak 1.622 ulasan positif dan negatif digunakan untuk membangun model Bi-LSTM pada masing-masing topik dengan penanganan ketidakseimbangan kelas menggunakan class weight. Model pada topik Penggunaan Layanan Maskapai Garuda Indonesia memperoleh accuracy 82,14% dan F1-score 68,35%, sedangkan topik Fitur Login dan Akses Akun memperoleh accuracy 90,32% dan F1-score 30,77%. Hasil penelitian menunjukkan bahwa sentimen pengguna didominasi oleh ulasan negatif, sehingga aspek layanan dan fitur aplikasi masih memerlukan perbaikan.
================================================================================================================================
As an archipelagic country, Indonesia experiences a continuous increase in the demand for air transportation each year. This condition highlights the important role of the aviation industry in supporting mobility and national economic growth. Garuda Indonesia, as the national airline and a member of the SkyTeam global alliance, continues to improve its digital services through the FlyGaruda application. Although the application has been downloaded more than one million times on Google Play Store, it still receives numerous complaints from users. Therefore, this study aims to analyze users' perceptions of the FlyGaruda application through sentiment analysis and topic modeling. Sentiment analysis was conducted using the InSet lexicon approach and the Bidirectional Long Short-Term Memory (Bi-LSTM) method, while topic modeling was performed using Latent Dirichlet Allocation (LDA). The topic modeling results identified two optimal topics with a coherence score of 0.4802, namely Garuda Indonesia Airline Service Usage and Login Features and Account Access in FlyGaruda. Sentiment labeling of 1,685 reviews produced 296 (17.6%) positive reviews, 1,326 (78.7%) negative reviews, and 63 (3.7%) neutral reviews. A total of 1,622 positive and negative reviews were used to develop Bi-LSTM models for each topic, with class imbalance handled using the class weight technique. The model for the Garuda Indonesia Airline Service Usage topic achieved an accuracy of 82.14% and an F1-score of 68.35%, while the model for the Login Features and Account Access in FlyGaruda topic achieved an accuracy of 90.32% and an F1-score of 30.77%. The findings indicate that users' perceptions of the FlyGaruda application are predominantly negative, suggesting that improvements in the application's services and features are still needed.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kata kunci: Analisis Sentimen, Bi-LSTM, FlyGaruda, Pemodelan Topik, LDA, Sentiment Analysis, Topic Modeling, LDA, Bi-LSTM, FlyGaruda
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Khurotaayun Pesona Subagja
Date Deposited:	02 Aug 2026 19:40
Last Modified:	02 Aug 2026 19:40
URI:	http://repository.its.ac.id/id/eprint/141828

Actions (login required)

View Item