Febrian, Giffani Rizky (2026) Monitoring Kualitas Layanan PELNI Mobile Menggunakan Diagram Kendali Demerit Berdasarkan Hasil Klasifikasi Multi-Label Dimensi Keluhan Pengguna Berbasis Modified TAM. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5003221014-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5003221014-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (11MB) | Request a copy |
Abstract
Perkembangan aplikasi digital pada sektor transportasi menjadikan ulasan pengguna sebagai sumber informasi penting untuk mengevaluasi kualitas layanan. Namun, meningkatnya volume data serta karakteristik keluhan yang dapat muncul pada lebih dari satu dimensi secara bersamaan (multi-label) menyebabkan analisis manual menjadi kurang efektif. Penelitian ini bertujuan mengembangkan sistem monitoring kualitas layanan aplikasi PELNI Mobile melalui integrasi metode Bi-LSTM Attention untuk klasifikasi multi-label dan Diagram Kendali Laney Demerit untuk monitoring kualitas layanan secara statistik. Data penelitian berupa ulasan pengguna aplikasi PELNI Mobile di Google Play Store periode Januari 2023 hingga Mei 2026 yang diklasifikasikan ke dalam dimensi Modified Technology Acceptance Model (Modified TAM), yaitu Reliability, Usefulness, Ease of Use, dan Non-Keluhan. Kinerja model Bi-LSTM Attention dibandingkan dengan LSTM dan Support Vector Machine (SVM), kemudian hasil klasifikasi dikonversi menjadi skor demerit untuk membangun Diagram Kendali Demerit Standar dan Diagram Kendali Laney Demerit. Hasil penelitian menunjukkan bahwa Bi-LSTM Attention memberikan performa terbaik dengan Subset Accuracy sebesar 79,94%, Macro F1-Score sebesar 79,10%, dan Macro ROC-AUC sebesar 94,67%, yang menunjukkan kemampuan klasifikasi yang baik meskipun data memiliki karakteristik class imbalance. Diagram Kendali Laney Demerit terbukti lebih efektif dibandingkan Diagram Kendali Demerit Standar dalam mengakomodasi fenomena underdispersion akibat variasi ukuran subgrup mingguan. Sistem monitoring berhasil mendeteksi tiga kondisi out-of-control pada periode pembentukan baseline dan satu kondisi out-of-control pada periode monitoring aktual. Analisis diagnostik menggunakan Diagram Pareto, N-gram, dan Word Cloud menunjukkan bahwa dimensi Reliability menjadi penyebab utama keluhan pengguna, terutama terkait permasalahan login, server, error, dan pembelian tiket. Hasil penelitian menunjukkan bahwa integrasi Bi-LSTM Attention dan Diagram Kendali Laney Demerit mampu mendukung monitoring kualitas layanan secara otomatis sekaligus menyediakan informasi diagnostik sebagai dasar evaluasi dan perbaikan layanan.
=======================================================================================================================================
The development of digital applications in the transportation sector has made user reviews an important source of information for evaluating service quality. However, manual analysis has become less effective due to the increasing volume of data and the multi-label nature of user complaints. This study aims to develop a service quality monitoring system for the PELNI Mobile application by integrating the Bidirectional Long Short-Term Memory with Attention (Bi-LSTM Attention) model for multi-label classification and the Laney Demerit Control Chart for statistical service quality monitoring. User reviews collected from the Google Play Store between January 2023 and May 2026 were categorized into the Modified Technology Acceptance Model (Modified TAM) dimensions: Reliability, Usefulness, Ease of Use, and Non-Complaint. The proposed model was compared with Long Short-Term Memory (LSTM) and Support Vector Machine (SVM), and the classification results were transformed into demerit scores to construct the Standard Demerit and Laney Demerit Control Charts. The Bi-LSTM Attention model achieved the best performance with a Subset Accuracy of 79.94%, a Macro F1-Score of 79.10%, and a Macro ROC-AUC of 94.67%, demonstrating robust classification performance despite the class imbalance characteristics of the dataset. Furthermore, the Laney Demerit Control Chart outperformed the Standard Demerit Control Chart in accommodating underdispersion caused by variations in weekly subgroup sizes. The monitoring system detected three out-of-control signals during the baseline establishment period and one during the actual monitoring period. Diagnostic analysis using Pareto Charts, N-grams, and Word Clouds identified the Reliability dimension as the primary source of user complaints, mainly related to login failures, server issues, application errors, and ticket purchasing. Overall, the integration of the Bi-LSTM Attention model and the Laney Demerit Control Chart provides an effective approach for automated service quality monitoring while supporting diagnostic analysis and continuous service quality improvement.
Actions (login required)
 |
View Item |