Analisis Sentimen Berbasis Aspek pada Layanan Transjakarta Melalui Media Sosial X dengan Metode Span – Aspect Sentiment Triplet Extraction

Arsyada, Muhammad Farrih Mahabbataka (2025) Analisis Sentimen Berbasis Aspek pada Layanan Transjakarta Melalui Media Sosial X dengan Metode Span – Aspect Sentiment Triplet Extraction. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211080-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy

Abstract

Transportasi massal seperti Transjakarta memegang peranan krusial dalam mobilitas perkotaan, namun berbagai keluhan terkait layanannya masih sering muncul. Penelitian ini mengembangkan Aspect-Based Sentiment Analysis (ABSA) menggunakan pendekatan Span-ASTE untuk menganalisis sentimen publik terhadap layanan Transjakarta secara lebih mendalam, dengan mengidentifikasi triplet (aspek, opini, sentimen) secara simultan dari data media sosial X. Data tweet berbahasa Indonesia dikumpulkan melalui web scraping disepanjang tahun 2024, kemudian diproses dan dianotasi secara manual menjadi korpus spesifik berisi 3.676 triplet.
Proses seleksi yang dilakukan terhadap empat pre-trained language model, IndoBERTweet teridentifikasi sebagai backbone optimal dengan F1-score 0.55 setelah melalui tahap tuning parameter. Implementasi IndoBERTweet pada dataset dengan kompleksitas berbeda menghasilkan kinerja F1-score 0.62 untuk dataset kompleksitas rendah dan 0.56 untuk dataset kompleksitas tinggi. Pada level subtugas individual untuk dataset kompleksitas tinggi, model mencapai F1-score 0.94 untuk klasifikasi sentimen, 0.78 untuk Aspect Term Extraction (ATE), dan 0.76 untuk Opinion Term Extraction (OTE). Proof of Concept berhasil mengimplementasikan pipeline lengkap mulai dari pengumpulan data mentah, ekstraksi triplet, hingga generalisasi aspek ke dalam 10 kategori utama. Output yang dihasilkan berupa peta sentimen terstruktur yang mengidentifikasi isu prioritas seperti ketidaktepatan jadwal dan keterbatasan armada, menyediakan dasar empiris untuk perbaikan layanan Transjakarta. Temuan ini tidak hanya membuktikan kemampuan model dalam menangani data noisy, tetapi juga memberikan rekomendasi praktis bagi perbaikan layanan Transjakarta, sekaligus menjadi terobosan dalam penerapan Span-ASTE untuk analisis lebih lanjut pada domain transportasi publik di Indonesia.
================================================================================================================================
Mass transportation systems like Transjakarta play a crucial role in urban mobility, yet service-related complaints remain frequent. This study develops an Aspect-Based Sentiment Analysis (ABSA) using the Span-ASTE approach to conduct an in-depth analysis of public sentiment toward Transjakarta services by simultaneously identifying triplets (aspect, opinion, sentiment) from social media data on platform X. Indonesian-language tweets were collected through web scraping throughout 2024, then processed and manually annotated into a specialized corpus containing 3,676 triplets. Through rigorous evaluation of four pre-trained language models, IndoBERTweet emerged as the optimal backbone architecture, achieving an F1-score of 0.55 after parameter tuning. When implemented on datasets of varying complexity, the model demonstrated differentiated performance with F1-score of 0.62 for low-complexity datasets and 0.56 for high-complexity datasets. At the individual subtask level for high-complexity data, the model achieved F1-scores of 0.94 for sentiment classification, 0.78 for Aspect Term Extraction (ATE), and 0.76 for Opinion Term Extraction (OTE).
The Proof of Concept successfully implemented a comprehensive pipeline from raw data collection and triplet extraction to aspect generalization across 10 main categories. The output yielded a structured sentiment map identifying priority issues such as schedule inaccuracy and fleet limitations, providing empirical evidence for Transjakarta service improvements. These findings not only demonstrate the model's capability in handling noisy user-generated content but also deliver actionable recommendations for Transjakarta while breaking new ground in Span-ASTE applications for public transportation analysis in Indonesia.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	ABSA, Span-ASTE, Transjakarta, Media Sosial X, IndoBERT
Subjects:	H Social Sciences > HE Transportation and Communications > HE147.6 Transportation--Planning Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Farrih Mahabbataka Arsyada
Date Deposited:	24 Jul 2025 04:07
Last Modified:	24 Jul 2025 04:07
URI:	http://repository.its.ac.id/id/eprint/120986

Actions (login required)

View Item