Container Handling Activities: A Forecasting Tool Proposal Using Wavelet and Convolutional Neural Networks with a Case Study of Turkey

Maritime transportation and ports are essential to global trade. Port operations, supported by standardized handling equipment, enhance multimodal transport and accelerate international commerce. With its strategic location and being surrounded by seas on three sides, Turkey holds a geographical advantage in global trade. Its ports---especially Ambarlı, Kocaeli, and Mersin---are the busiest in terms of container handling. The COVID-19 pandemic disrupted container operations worldwide, including in Turkey. These disruptions highlighted the need for proactive strategies to maintain port efficiency during future crises. This study utilizes monthly container data (2019--2024) from three major Turkish ports to forecast container handling throughput using machine and deep learning methods: convolutional neural networks (CNN) and wavelet neural networks (WNN). The study presents a forecasting analysis for the year 2025, incorporating both observed data (January--May) and projected data (June--December), and compares the accuracy and robustness of the two techniques. The results show that both CNN and WNN offer promising accuracy, but CNN outperforms WNN in handling complex container data and provides more robust predictions. These methods support strategic planning, crisis management, investment decisions, and competitive advantage in port operations. In this context, the developed forecasting models have the potential to enhance the resilience of Turkish ports within global supply chains and help them become better prepared for future uncertainties.