Precipitation forecasting is vital for managing disasters, urban traffic, and agriculture. This study develops an improved model for short-term precipitation forecasting by combining Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Gated Recurrent Unit (GRU). Using precipitation data from January 1, 2019, to December 31, 2022, as a sample, the model capitalizes on CEEMDAN’s superior signal decomposition capabilities and GRU’s ability to capture nonlinear dynamic patterns in time series. To assess the model’s effectiveness, comparisons were conducted with 12 benchmark models, including CEEMDAN-LSTM, EMD-GRU, EMD-LSTM, BI-LSTM, GRU, LSTM, and TCN. The results demonstrate that the CEEMDAN-GRU model achieves higher accuracy and stability in short-term precipitation forecasting. Leveraging an Adam optimizer with adaptive learning rate reduction enhances convergence and ensures reliable predictions, achieving an R²of 0.7915, MAE of 0.05382, and MSE of 0.09081.

Based on Automatic Identification System (AIS) data in maritime settings, this paper explores the limitations of traditional Recurrent Neural Networks in extracting features from complex vessel trajectory sequences. We propose a fusion model, namely CNN-RGRU-Attention, for vessel trajectory prediction. The model integrates Convolutional Neural Network (CNN), Attention Mechanism, and Gated Recurrent Unit (GRU). The effectiveness of the model is validated using actual AIS data, demonstrating significant improvements in metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). The CNN-RGRU-Attention model provides crucial theoretical support for enhancing the safety management of maritime traffic services.

Federated learning enables multiple nodes to perform local computations and collaborate to complete machine learning tasks without centralizing private data of nodes. However, the frequent model gradients upload/download operations required by the framework result in high communication costs, which have become the main bottleneck for federated learning as deep models scale up, hindering its performance. In this paper, we propose a two-layer accumulated quantized compression algorithm (TLAQC) that effectively reduces the communication cost of federated learning. TLAQC achieves this by reducing both the cost of individual communication and the number of global communication rounds. TLAQC introduces a revised quantization method called RQSGD, which employs zero-value correction to mitigate ineffective quantization phenomena and minimize average quantization errors. Additionally, TLAQC reduces the frequency of gradient information uploads through an adaptive threshold and parameter self-inspection mechanism, further reducing communication costs. It also accumulates quantization errors and retained weight deltas to compensate for gradient knowledge loss. Through quantization correction and two-layer accumulation, TLAQC significantly reduces precision loss caused by communication compression. Experimental results demonstrate that RQSGD achieves an incidence of ineffective quantization as low as 0.003% and reduces the average quantization error to 1.6 × $${10}^{-5}$$10 - 5 . Compared to full-precision FedAVG, TLAQC compresses uploaded traffic to only 6.73% while increasing accuracy by 1.25%.

In today’s time, maritime transport is becoming the mainstream. Ports have the problems of inefficient berthing and unreasonable allocation of shore and bridge resources. In this paper, we propose a model to predict the ship arrival time based on AIS data and trajectory inflection points to predict the anchor arrival time, aiming to solve the problems of low berthing efficiency and unreasonable allocation of shore and bridge resources. Experiments show that the minimum prediction error of the model is 8min, the maximum error is 1h, and the average error is 30min; and compared with the ship schedule data, the maximum error is 7 days, the minimum error is 1 day, and the average error is 2.75 days, so the time got from this model is better than the ship schedule, which can effectively improve the berthing efficiency of the port and the reasonable allocation of shore and bridge resources. The model has good accuracy and effectiveness.

With the development of the maritime economy and the widespread adoption of the Automatic Identification System (AIS), the number of vessels at sea has surged. Unfortunately, some vessels illicitly alter their Maritime Mobile Service Identity (MMSI) codes for personal gain, posing challenges to port management. Currently, the detection of “Multiple MMSI on a Single Vessel” vessels still relies on manual inspections by relevant authorities. This paper proposes a “Multiple MMSI on a Single Vessel” detection model based on Automatic Identification System (AIS) datas. It employs interval sampling combined with cosine similarity to enhance the accuracy and efficiency of identifying such vessels. This technology finds broad applications in port management, safety, vessel regulation, and maritime law enforcement. The paper sets the similarity threshold at 0.7, a distance threshold at 2.7 n mile, and a time threshold at 3 s. Experimental results demonstrate that the model achieves an average detection accuracy of 92.25%. Therefore, the proposed model effectively and accurately detects “Multiple MMSI on a Single Vessel” illegal vessels, significantly enhancing the efficiency of port and maritime management authorities. This model proves to be a rational and efficient solution.

Strawberry quality grading helps producers to better manage inventory, transportation and sales, and improve product market competitiveness. Currently, this work is mostly carried out by manual grading, which is not only inefficient, but also relies on the experience of the grader, which is easy to cause grading errors. Aiming at the complex background information present in strawberry images, a grading fusion model is proposed, which adopts YOLOv5 target detection algorithm in the first level to recognize the strawberry in its natural state and crop its background information; In the second level, a stacking fusion model is adopted, which combines the neural network classification model and the machine learning classifier to realize the lossless quality grading of strawberry; The loss function is improved by introducing the Label Smoothing method, which makes the model more suitable for strawberry grading this kind of non-standard product classification task; Comparison experimental results show that the accuracy of the strawberry grading fusion model based on complex background images proposed in this paper reaches 85.4%, which improves the classification accuracy by 13% comparing with that of the single-stage model.