Prediksi Peluang Tubrukan Kapal Melalui Implementasi Hybrid Bayesian Network Menggunakan Algoritma Dynamic Discretization

Zahabiya, Renata (2023) Prediksi Peluang Tubrukan Kapal Melalui Implementasi Hybrid Bayesian Network Menggunakan Algoritma Dynamic Discretization. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043201107_Undergraduate Thesis.pdf - Accepted Version Restricted to Repository staff only until April 2026. Download (4MB) | Request a copy

Abstract

Tubrukan kapal merupakan jenis kecelakaan pelayaran dengan dampak yang tidak sedikit, meliputi kerusakan pada kapal terkait dan kapal lain di sekitarnya, membahayakan lingkungan laut karena tumpahan minyak, serta dapat memicu terjadinya kecelakaan pelayaran lain. Terdapat beberapa faktor penyebab terjadinya tubrukan kapal, seperti faktor teknis, faktor cuaca, dan faktor manusia. Oleh karena itu, prediksi tubrukan kapal melalui identifikasi faktor yang menyebabkannya perlu dilakukan sebagai langkah awal dalam mencegah dan mengurangi terjadinya tubrukan kapal. Hybrid Bayesian Network merupakan metode machine learning yang dapat digunakan dalam melakukan prediksi tubrukan kapal. Metode ini menjelaskan struktur hubungan antara berbagai variabel random yang kompleks ke dalam bentuk diagram berdasarkan teori peluang bersyarat yang tidak hanya mempertimbangkan variabel kategorik, tetapi juga mempertimbangkan variabel kontinu. Prediksi tubrukan kapal melalui implementasi Hybrid Bayesian Network akan dilakukan menggunakan algoritma dynamic discretization yang memiliki keunggulan dibandingkan static discretization. Model prediksi peluang tubrukan kapal di Indonesia yang didapatkan melalui implementasi Hybrid Bayesian Network menggunakan algoritma dynamic discretization menghasilkan nilai peluang tubrukan kapal sebesar 0,67 serta nilai akurasi sebesar 94,74%, dimana faktor yang paling berkontribusi terhadap terjadinya tubrukan kapal adalah good seamanship, pengambilan keputusan, preventive timing, kegagalan mesin, serta kemampuan manuver kapal.
=================================================================================================================================
Ship collision is a type of shipping accident with a lot of impact, including damage to related ships and other ships in the vicinity, endanger the marine environment due to oil spills, and can trigger other shipping accidents. There are several factors that contribute to causing ship collisions, such as technical factors, weather factors, and human factors. Therefore, the prediction of ship collisions through the identification of the factors that cause it needs to be done as a initial step in preventing and reducing ship collisions. Hybrid Bayesian Network is a machine learning method that can be used to predict ship collisions. This method explains the structure of relationships between various complex random variables in the form of diagrams based on conditional probability theory which not only consider categoric variables, but also considers continuous variables. Ship collision prediction through the implementation of Hybrid Bayesian Network will be carried out using a dynamic discretization algorithm which has advantages over static discretization. The prediction model for ship collisions probability in Indonesia obtained through the implementation of Hybrid Bayesian Network using a dynamic discretization algorithm produces a ship collision probability value of 0,67 and an accuracy value of 94,74%, where the factors that most contribute to the occurrence of ship collisions are good seamanship, decision making, preventive timing, technical failure, and maneuverability.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Dynamic Discretization, Hybrid Bayesian Network, Tubrukan Kapal, Ship Collision
Subjects:	Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory.
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Renata Zahabiya
Date Deposited:	18 Mar 2024 02:11
Last Modified:	18 Mar 2024 02:11
URI:	http://repository.its.ac.id/id/eprint/107821

Actions (login required)

View Item