Perancangan Sistem Rekomendasi Emergency Reponse Procedure Kapal Dengan Metode Natural Language Processing

Faishal, Erza Multazami (2025) Perancangan Sistem Rekomendasi Emergency Reponse Procedure Kapal Dengan Metode Natural Language Processing. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5018211005_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Kecelakaan kapal merupakan peristiwa yang dapat menyebabkan dampak besar terhadap kondisi finansial maupun korban jiwa. Data menunjukkan bahwa banyak kecelakaan terjadi akibat kesalahan manusia, termasuk kurangnya informasi dan pelatihan yang memadai untuk merespons keadaan darurat. Oleh karena itu, penelitian ini bertujuan untuk merancang sistem rekomendasi Emergency Response Procedure (ERP) kapal berbasis Natural Language Processing (NLP) untuk membantu kru kapal dalam mengambil tindakan penyelamatan yang cepat dan tepat. Penelitian ini mencakup perancangan dataset NLP yang komprehensif, dikumpulkan dari Australian Maritime Safety Authority (AMSA), SQE Marine, dan Modul Prosedur Darurat & SAR Indonesia. Dataset ini diperkaya melalui augmentasi variasi kalimat dan pemetaan sinonim untuk memastikan keragaman linguistik. Sistem rekomendasi dirancang dengan arsitektur hibrida, mengintegrasikan Large Language Model (LLM) berbasis Transformer (TinyLlama) untuk pemahaman kontekstual dan generasi SOP dinamis, TF-IDF untuk analisis kesamaan teks, serta pendekatan berbasis aturan dan pola. Model ini mampu mengidentifikasi lima kategori situasi darurat utama: kebakaran, tubrukan, kandas, kebocoran, dan man overboard. Hasil evaluasi model menunjukkan kinerja yang optimal dengan akurasi 88.89%, F1-Score 88.15%, presisi 92.59%, dan recall 88.89%. Pengujian sistem melalui berbagai skenario darurat menunjukkan kemampuan model dalam mengolah input pengguna dan memberikan rekomendasi prosedur yang relevan. Perancangan user interface juga menjadi bagian integral untuk memastikan kemudahan penggunaan. Penelitian ini berkontribusi pada pengembangan teknologi di bidang perkapalan dan diharapkan dapat membantu kru kapal dalam melakukan tindakan penyelamatan saat terjadi kecelakaan.
=======================================================================================================================
Ship accidents can cause significant financial and human casualties. Data indicates that many accidents result from human error, including insufficient information and inadequate training for emergency response. Therefore, this research aims to design a ship Emergency Response Procedure (ERP) recommendation system based on Natural Language Processing (NLP) to assist ship crews in taking swift and appropriate rescue actions. This study involves the design of a comprehensive NLP dataset, collected from the Australian Maritime Safety Authority (AMSA), SQE Marine, and the Indonesian Emergency and SAR Procedure Module. The dataset is enriched through sentence variation augmentation and synonym mapping to ensure linguistic diversity. The recommendation system is designed with a hybrid architecture, integrating a Transformer-based Large Language Model (LLM) (TinyLlama) for contextual understanding and dynamic SOP generation, TF-IDF for text similarity analysis, and rule-based and pattern matching approaches. This model can identify five main emergency situations: fire, collision, grounding, leakage, and man overboard. The model evaluation results show optimal performance with an accuracy of 88.89%, an F1-Score of 88.15%, precision of 92.59%, and recall of 88.89%. System testing through various emergency scenarios demonstrates the model's ability to process user input and provide relevant procedural recommendations. User interface design was also an integral part to ensure ease of use. This research contributes to the development of technology in the shipping industry and is expected to assist ship crews in performing rescue actions during accidents.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Sistem Rekomendasi, Emergency Response Procedure, NLP, Machine Learning, Kapal. Recommendation System, Emergency Response Procedure, NLP, Machine Learning, Ship.
Subjects:	V Naval Science > VK > VK1258 Ships--Fires and fire prevention V Naval Science > VK > VK200 Merchant marine--Safety measures
Divisions:	Faculty of Marine Technology (MARTECH) > Naval Architecture and Shipbuilding Engineering > 36201-(S1) Undergraduate Thesis
Depositing User:	Erza Multazami Faishal
Date Deposited:	04 Aug 2025 06:59
Last Modified:	04 Aug 2025 06:59
URI:	http://repository.its.ac.id/id/eprint/126253

Actions (login required)

View Item