Perbandingan Algoritma AI untuk Sistem Pendukung Keputusan Penjadwalan Bongkar Muat Barang di Pelabuhan Semayang Balikpapan

Abstract

Optimizing loading and unloading scheduling at Semayang Port, Balikpapan, is a complex challenge that requires an artificial intelligence approach. This study compares the effectiveness of the Genetic Algorithm (GA) and Neural Network (NN) as decision support systems in minimizing dwell time and improving operational efficiency. The research data covers the period 2020–2023, including ship arrival times, cargo types and volumes, equipment availability, and external factors such as weather and tides. GA is implemented using a chromosomal representation based on ship priority and fitness functions to optimize dwell time and dock utilization. NN employs a Long Short-Term Memory (LSTM) architecture to process port time-series data. The evaluation of three scenarios—peak, normal, and low seasons—shows that GA reduces dwell time by 22.3% compared to conventional methods, while NN predicts potential delays with an accuracy of 87.2%. GA performs best in static scheduling, while NN is effective in handling dynamic uncertainties. The hybrid GA-NN solution provides a balance between performance and complexity, with the potential to reduce operational costs by up to IDR 1.2 billion per year. This study presents a framework for the comparative evaluation of AI algorithms in port logistics and offers practical recommendations for optimizing operations at Semayang Port.