Teguh, Saladin (2026) Pengembangan Algoritma Particle Swarm Optimization Untuk Menyelesaikan Berth Allocation Problem Berdasarkan Analisis Waktu Kedatangan Kapal Menggunakan Metode Deep Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 6010241002_Master_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
6010241002_Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy |
Abstract
Efisiensi operasional pelabuhan sangat dipengaruhi oleh ketepatan penjadwalan tambat kapal. Pada Continuous Berth Allocation Problem (CBAP), ketidakpastian Estimated Time of Arrival (ETA) dapat meningkatkan waiting time, menimbulkan konflik penjadwalan, dan menurunkan utilisasi dermaga. Penelitian ini bertujuan mengembangkan model prediksi ETA berbasis data Automatic Identification System (AIS) menggunakan Deep Stacked Long Short-Term Memory (Deep Stacked LSTM), mengintegrasikan hasil prediksi tersebut ke dalam optimasi CBAP menggunakan Particle Swarm Optimization (PSO) yang dimodifikasi, serta mengevaluasi kinerja model usulan. Penelitian menggunakan 169.031 data AIS valid yang merepresentasikan 443 voyage di Terminal Jamrud, Pelabuhan Tanjung Perak. Data diproses melalui preprocessing, feature engineering, normalisasi Min-Max Scaling, pembentukan sliding window, prediksi ETA menggunakan Deep Stacked LSTM, dan optimasi CBAP menggunakan PSO yang dimodifikasi dengan strategi Shortest Processing Time, Heuristic Seeding, dan Emergency Overrule. Hasil penelitian menunjukkan bahwa model usulan menurunkan Average Waiting Time dari 31,80 menjadi 22,76 jam, Maximum Waiting Time dari 109,30 menjadi 80,20 jam, serta makespan dari 24.721 menjadi 20.716 jam. Jumlah kapal yang mengalami antrean lebih dari 72 jam berkurang dari 34 menjadi 10 kapal tanpa berth overlap. Uji Wilcoxon Signed-Rank menghasilkan p-value < 0,001, yang menunjukkan peningkatan performa signifikan. Model yang diusulkan berpotensi mendukung pengambilan keputusan penjadwalan tambat kapal secara lebih efisien menuju implementasi Smart Port.
=================================================================================================================================
Port operational efficiency is highly dependent on effective berth scheduling. In the Continuous Berth Allocation Problem (CBAP), uncertainty in the Estimated Time of Arrival (ETA) may increase vessel waiting time, create scheduling conflicts, and reduce berth utilization. This study aims to develop an ETA prediction model based on Automatic Identification System (AIS) data using a Deep Stacked Long Short-Term Memory (Deep Stacked LSTM) network, integrate the predicted ETA into a CBAP optimization model using a modified Particle Swarm Optimization (PSO) algorithm, and evaluate the performance of the proposed model. The study employed 169,031 valid AIS records representing 443 vessel voyages at Jamrud Terminal, Tanjung Perak Port. The methodology consisted of data preprocessing, feature engineering, Min-Max normalization, sliding window generation, ETA prediction using the Deep Stacked LSTM model, and CBAP optimization using a modified PSO incorporating the Shortest Processing Time, Heuristic Seeding, and Emergency Overrule strategies. The results demonstrate that the proposed model reduced the Average Waiting Time from 31.80 to 22.76 hours, the Maximum Waiting Time from 109.30 to 80.20 hours, and the makespan from 24,721 to 20,716 hours. The number of vessels experiencing waiting times exceeding 72 hours decreased from 34 to 10, while berth overlaps were completely eliminated. Furthermore, the Wilcoxon Signed-Rank Test yielded a p-value of < 0.001, confirming a statistically significant improvement in scheduling performance. The proposed framework has the potential to support more efficient berth scheduling and contribute to the implementation of Smart Port decision-support systems.
Actions (login required)
 |
View Item |