Accurate prediction of vessel estimated time of arrival (ETA) requires both reliable route reconstruction and efficient moving-time estimation. However, conventional maritime traffic networks (MTNs) often suffer from excessive network complexity and may inadequately represent heterogeneous vessel movements. This paper proposes a lightweight MTN (LMTN) framework that combines adaptive sampling, dual-feature spatial clustering of applications with noise (DF-SCAN), and two-stage sequential prediction (TSSP). Adaptive sampling selectively retains operationally informative automatic identification system (AIS) nodes, whereas DF-SCAN considers spatial proximity and navigational characteristics to construct representative route networks. TSSP incorporates voyage-level speed over ground information into segment-level moving-time estimation. Using largescale real-world AIS data, the proposed LMTN reduced network representation by 89.38% while maintaining competitive route-reconstruction performance and achieving zero route-generation failures in the evaluated trajectories. The average time required for route searching and ETA calculation was 0.064 s, compared to 1.519 s using dense H3 representation. ETA experiments showed that the LMTN-based prediction became more effective with increasing voyage distance; moreover, TSSP improved prediction accuracy under several mid- to long- range conditions. Experiments on various vessel types provided preliminary evidence of cross-vessel applicability, whereas supplementary deep-learning benchmarks provided further comparative context. Overall, the results demonstrate that lightweight route representation and hierarchical speed-informed moving-time estimation can support efficient vessel ETA prediction; moreover, the resulting route, speed, and ETA information can serve as supplementary decision-support inputs for vessel-arrival monitoring, berth planning, and terminal-resource coordination.

Constructing health indicators (HIs) that quantitatively represent the state of a machine is essential for the condition-based maintenance of rotating machinery. Traditional physics-based HI construction methods have limited effectiveness; for example, virtual HIs integrating multiple features have limited applicability because of the manual feature extraction process that depends on expert knowledge and lack of model adaptability. Existing studies focus on single domains (raw signals) or use time–frequency domain spectrograms and fail to adequately extract meaningful frequency band patterns extended along specific axes. These limitations in representation capacity lead to a degraded performance in HI extraction. This paper proposes a novel deep learning-based HI extraction method that uses a supervised time–frequency feature fusion autoencoder for fusing multidomain features. The proposed model includes spatiotemporal feature extraction from raw signals using a multiscale convolution neural network Transformer block, extraction of temporally resolved frequency features from time–frequency maps using rectangular kernels and cross-attention mechanisms, fusion of latent representations to extract one-dimensional HI, and simultaneous reconstruction of the input signal from the extracted HI unified encoder–decoder architecture. During training, the model optimizes a composite loss function that comprises a shape constraint function and domain-specific reconstruction errors using a dynamic weight-averaging strategy. The proposed method is validated using publicly available rotating machinery datasets and demonstrates superior performances across four evaluation metrics compared with those of existing methods. This study highlights the potential of automated high-performance HI extraction for rotating machinery and offers practical contributions to industrial maintenance applications.

Import container dwell time (ICDT) prediction is a key task for improving productivity in container terminals, as accurate predictions enable the reduction of container re-handling operations by yard cranes. Achieving this objective requires accurately predicting the dwell time of individual containers. However, the primary determinants of dwell time-owner information and cargo information-are recorded as unstructured text, which limits their effective use in machine learning models. This study addresses this limitation by proposing a collaborative framework that integrates generative artificial intelligence (Gen AI) with machine learning. The proposed framework employs Gen AI to standardize unstructured information into standard international codes, with dynamic re-prediction triggered by electronic data interchange state updates, enabling the machine learning model to predict ICDT accurately. Extensive experiments conducted on real container terminal data demonstrate that the proposed methodology achieves a 13.88% improvement in mean absolute error compared to conventional models that do not utilize standardized information. Furthermore, applying the improved predictions to container stacking strategies achieves up to 14.68% reduction in the number of relocations, thereby empirically validating the potential of Gen AI to enhance productivity in container terminal operations. Overall, this study provides both technical and methodological insights into the adoption of Gen AI in port logistics and its effectiveness.

Maritime transport handles over 80% of global trade but rising traffic density and congestion have increased collision risk. Traditional assessment methods, including the ship domain and closest point-of-approach approaches, are limited to pairwise interactions and cannot capture regional traffic dynamics. To overcome these shortcomings, we propose a regional collision risk assessment framework integrating AIS-derived relative traffic representations, convolutional neural networks, and explainable artificial intelligence. A gradient-weighted class activation mapping framework generated risk influence distribution maps and a centroid radial distribution, while a novel indicator, the risk influence radius (RIR), quantified radial influence patterns. Experiments with 3,660 low-risk and 366 high-risk cases from South Korean coastal waters revealed that the Visual Geometry Group (VGG) network backbone achieved the highest performance, with an average F1-score of 0.9324, outperforming ResNeXt, EfficientNet, ResNet, and MobileNetV3 models. Statistical tests revealed a significant difference in RIR (p < 0.05) between high- and low-risk collision areas at 10 and 20 km, but not at 30 km. These results confirm RIR as an interpretable indicator of collision risk at regional scales, while highlighting its limitations at broader ranges. The framework combines accuracy with interpretability and supports practical applications such as maritime monitoring, route planning, and early warning systems.applications such as maritime monitoring, route planning, and early warning systems.

As global supply chains grow increasingly complex and cargo volumes continue to rise, the accurate prediction of vessel turnaround time (VTT) has become essential for enhancing port operational efficiency. Previous studies primarily utilized static features, such as vessel specifications and container volumes, often failing to capture the dynamic nature of real-world operations. To address this limitation, we propose a forecasting framework that incorporates dynamic operation indicators (OIs) from a two-stage queuing model into time-series models. Leveraging operational data from Busan Port, our approach achieves of up to 28% improvement in mean absolute percentage error compared with existing techniques. Explainability analysis using Shapley additive explanations (SHAP) further demonstrates that OIs rank among the most influential features. By providing more reliable VTT predictions, port operators can enhance berth planning, optimize resource allocation, minimize scheduling delays, and enhance overall port efficiency. These results highlight the practical value of the framework for data-driven decision-making in port operations.

Forecasting the maritime economics index, including container volume and Baltic Panamax Index, is essential for long-term planning and decision-making in the shipping industry. However, studies on container volume prediction are not sufficient, and the bulk freight index has highly fluctuating characteristics, which pose a challenge in long-term prediction. This study proposes a new hybrid framework for the long-term prediction of the maritime economics index. The framework consists of time-series decomposition to break down a time-series into several components (trend, seasonality, and residual), a two-stage attention mechanism that prioritizes important variables to increase long-term prediction accuracy and a long short-term memory network that predicts and combines all components to derive the final predictive outcome. Extensive experiments are conducted using the container volume data, bulk freight index data, and various external variables. The proposed framework achieved a better predictive performance than existing time-series methods, including conventional machine learning and deep learning-based models, in the long-term prediction of container volume and the Baltic Panamax Index. Hence, the proposed method can help in decision-making through accurate long-term predictions of the maritime economics index.

Time-series forecasting (TSF) is a traditional problem in the field of artificial intelligence, and models such as recurrent neural network, long short-term memory, and gate recurrent units have contributed to improving its predictive accuracy. Furthermore, model structures have been proposed to combine time-series decomposition methods such as seasonal-trend decomposition using LOESS. However, this approach is learned in an independent model for each component, and therefore, it cannot learn the relationships between the time-series components. In this study, we propose a new neural architecture called a correlation recurrent unit (CRU) that can perform time-series decomposition within a neural cell and learn correlations (autocorrelation and correlation) between each decomposition component. The proposed neural architecture was evaluated through comparative experiments with previous studies using four univariate and four multivariate time-series datasets. The results showed that long- and short-term predictive performance was improved by more than 10%. The experimental results indicate that the proposed CRU is an excellent method for TSF problems compared to other neural architectures.