Devadatta, Gusti Putu Nayaka (2026) Analisis Sentimen Keluhan Pelanggan Berbasis Bert Dan Ishikawa Diagram Sebagai Dasar Pemetaan Masalah Serta Perbaikan Layanan Distribusi Air Pada Perumda Air Minum Tirta Mangutama. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5010221120-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5010221120-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (12MB) | Request a copy |
Abstract
Perkembangan media sosial di Indonesia telah menjadikan platform digital sebagai ruang utama bagi masyarakat untuk menyampaikan opini dan keluhan terhadap layanan publik, termasuk layanan distribusi air yang dikelola oleh Perusahaan Umum Daerah Air Minum (Perumda AM). Banyaknya komentar tidak terstruktur menyulitkan pemantauan persepsi publik secara manual. Penelitian ini bertujuan untuk menganalisis sentimen keluhan pelanggan secara otomatis dari komentar akun Instagram resmi Perumda Tirta Mangutama, Kabupaten Badung, Bali, serta mengidentifikasi akar penyebab keluhan menggunakan Ishikawa Diagram sebagai dasar rekomendasi perbaikan layanan distribusi air. Metode yang digunakan adalah analisis sentimen berbasis deep learning menggunakan model BERT, khususnya model IndoBERTweet. Sebanyak 739 komentar berhasil dikumpulkan melalui teknik web scraping dan menggunakan tiga metode pelabelan (LLM, Rule Based dan Lexicon Based, serta Manual Labelling) sebagai ground truth. Data kemudian diproses melalui tahapan preprocessing teks dan dibagi menjadi data latih dan data validasi dengan komposisi 80:20. Hasil klasifikasi sentimen negatif kemudian dibedah secara diagnostik menggunakan Ishikawa Diagram untuk memetakan akar masalah berdasarkan lima faktor, yaitu material, man, environment, machine, dan method. Hasil terbaik penelitian menunjukkan bahwa model IndoBERTweet berhasil mengklasifikasikan sentimen keluhan pelanggan dengan performa yang sangat baik, menghasilkan nilai accuracy sebesar 95,95%, weighted average precision 96,39%, weighted average recall 95,95%, dan weighted average F1-score sebesar 96,14%. Berdasarkan distribusi data, kelas sentimen negatif sangat mendominasi dengan 709 komentar (95,94%), diikuti oleh sentimen netral 18 komentar (2,44%), dan positif 12 komentar (1,62%). Berdasarkan klasifikasi diperoleh nilai NBR -96,67%, yang menunjukkan topik ini didominasi sentimen negatif. Hasil pemetaan menggunakan Ishikawa Diagram menunjukkan bahwa faktor teknis distribusi air pada kategori machine menjadi penyebab keluhan paling dominan, dengan keluhan gangguan suplai air atau air mati sebanyak 424 komentar, disusul kendala geografis dataran tinggi (environment) di wilayah Kutuh, Kampial, dan Pecatu sebanyak 66 komentar, serta ketidakjelasan estimasi informasi gangguan (method) sebanyak 68 komentar. Integrasi IndoBERTweet dan Ishikawa Diagram dalam penelitian ini berhasil memberikan rekomendasi prioritas perbaikan layanan yang terstruktur dan berbasis data untuk meningkatkan kualitas layanan distribusi air daerah.
===============================================================================================================================
The expansion of social media in Indonesia has transformed digital platforms into a primary space for the public to express opinions and complaints regarding public services, including water distribution utilities managed by regional water utility companies (Perumda AM). The massive volume of unstructured comments makes manual monitoring of public perception challenging. This study aims to automatically analyze customer complaint sentiments from the official Instagram comments of Perumda AM Tirta Mangutama, Badung Regency, Bali, and identify the root causes of these complaints using an Ishikawa Diagram as a basis for water distribution service improvement recommendations. The method employs deep learning-based sentiment analysis using the BERT model, specifically IndoBERTweet. A total of 739 comments were collected via web scraping techniques utilizing three labeling methods (LLM, Rule-Based and Lexicon-Based, as well as Manual Labeling) to establish the ground truth. The data were then processed through text preprocessing stages and partitioned into training and validation sets with an 80:20 ratio. The classification results of negative sentiments were further analyzed diagnostically using an Ishikawa Diagram to map the root problems based on five factors, material, man, environment, machine, and method. The optimal research results demonstrate that the IndoBERTweet model successfully classified customer complaints with excellent performance, achieving an accuracy of 95.95%, a weighted average precision of 96.39%, a weighted average recall of 95.95%, and a weighted average F1-score of 96.14%. Based on the data distribution, the negative sentiment class heavily dominated with 709 comments (95.94%), followed by neutral sentiment with 18 comments (2.44%), and positive sentiment with 12 comments (1.62%). Based on the classification, a Net Brand Reputation (NBR) value of -96.67% was obtained, indicating that this topic is overwhelmingly dominated by negative sentiment. The Ishikawa Diagram mapping revealed that technical water distribution factors within the machine category were the most dominant cause of complaints, primarily regarding water supply disruptions or outages with 424 comments, followed by geographic constraints of high-altitude areas (environment) in the Kutuh, Kampial, and Pecatu regions with 66 comments, and unclear disruption time estimates (method) with 68 comments. The integration of IndoBERTweet and the Ishikawa Diagram in this study successfully provides structured, data-driven priority recommendations to enhance the quality of regional water distribution services.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Analisis Sentimen, IndoBERT, Ishikawa Diagram, Keluhan Pelanggan,Distribusi Air. Sentiment Analysis, IndoBERT, Ishikawa Diagram, Customer Complaints, Water Distribution. |
| Subjects: | Q Science Q Science > Q Science (General) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) T Technology > T Technology (General) > T57.5 Data Processing |
| Divisions: | Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis |
| Depositing User: | Gusti Putu Nayaka Devadatta |
| Date Deposited: | 01 Aug 2026 02:11 |
| Last Modified: | 01 Aug 2026 02:11 |
| URI: | http://repository.its.ac.id/id/eprint/141784 |
Actions (login required)
 |
View Item |