Xinying Chen

This paper proposes the YOLOv8n_H method to address issues regarding parameter redundancy, slow inference speed, and suboptimal detection precision in contemporary helmet-wearing target recognition algorithms. The YOLOv8 C2f module is enhanced with a new SC_Bottleneck structure, incorporating the SCConv module, now termed SC_C2f, to mitigate model complexity and computational costs. Additionally, the original Detect structure is substituted with the PC-Head decoupling head, leading to a significant reduction in parameter count and an enhancement in model efficiency. Moreover, the original Detect structure is replaced by the PC-Head decoupling head, significantly reducing parameter count and enhancing model efficiency. Finally, regression accuracy and convergence speed are boosted by the dynamic non-monotonic focusing mechanism introduced through the WIoU boundary loss function. Experimental results on the expanded SHWD dataset demonstrate a 46.63% reduction in model volume, a 44.19% decrease in parameter count, a 54.88% reduction in computational load, and an improvement in mean Average Precision (mAP) to 93.8% compared to the original YOLOv8 algorithm. In comparison to other algorithms, the model proposed in this paper markedly reduces model size, parameter count, and computational load while ensuring superior detection accuracy.

To tackle the challenge of low accuracy in stock prediction within high-noise environments, this paper innovatively introduces the CED-PSO-StockNet time series model. Initially, the model decomposes raw stock data using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) technique and reconstructs the components by estimating their frequencies via the extreme point method. This process enhances component stability and mitigates noise interference. Subsequently, an Encoder-Decoder framework equipped with an attention mechanism is employed for precise prediction of the reconstructed components, facilitating more effective extraction and utilization of data features. Furthermore, this paper utilizes an Improved Particle Swarm Optimization (IPSO) algorithm to optimize the model parameters. On the Pudong Bank dataset, through ablation experiments and comparisons with baseline models, various optimization strategies incorporated into the proposed CED-PSO-StockNet model were effectively validated. Compared to the standalone LSTM model, CED-PSO-StockNet achieved a remarkable 45.59% improvement in the R² metric. To further assess the model’s generalization capability, this paper also conducted comparative experiments on the Ping An Bank dataset, and the results underscored the significant advantages of CED-PSO-StockNet in the domain of stock prediction.

Air pollution significantly impacts human health, the economy, and social stability. Accurate air quality prediction is crucial for effective prevention and control of air pollution. To address the highly nonlinear and non-stationary nature of PM2.5 sequences and the spatiotemporal dependencies between the sequences and air quality monitoring stations, a spatiotemporal hybrid prediction model based on CEEMDAN-GCN-GRU is proposed.First, PM2.5 sequences from each air quality monitoring station are decomposed using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to obtain multiple intrinsic mode functions (IMFs) and residual components (RES). Next, each component is combined with other features and input into the Graph Convolution Network (GCN) to capture the spatial correlations between PM2.5 concentrations at different monitoring stations. The combined decomposed PM2.5 sequence data and its spatial correlations are then used as input to the Gated Recurrent Unit (GRU) to extract spatiotemporal features. Finally, the individual sub-sequences output from the GRU are predicted using a linear regression layer, and the results are superimposed to obtain the final PM2.5 prediction results.Experiments conducted with 10 state-controlled air quality monitoring stations in Dalian city show that the proposed model outperforms comparative models in terms of root mean square error (RMSE) and mean absolute error (MAE), achieving reductions of 33.86% and 46.04%, respectively, compared to single benchmark LSTM and GRU models.

Currently, there are several issues with complicated road conditions, including significant variations in the size of the vehicle target, dense occlusion, high overlap rate, and inadequate lighting. The results of target detection are imprecise, prone to error, and misinterpreted. To address these issues, this paper proposes a technique for target detection using multi-scale fusion with an improved version of YOLOv8. Firstly, a new feature fusion network based on the weighted bidirectional feature network and the adaptive spatial feature fusion is proposed to strengthen the feature extraction and improve the visual representation ability. Secondly, to further improve the ability to detect small targets, a multi-scale target detection layer is incorporated. Meanwhile, an enhanced CIOU loss function is presented to accelerate the convergence of the model. Finally, this article combines ordinary convolution with depth separable convolution to improve the spatial feature information extraction capability and strengthen the model's generalization ability. Experiments were conducted on the KITTI dataset, and the results showed that the mAP of the improved network improved to 93.0%, which exceeded that of the traditional target detection network in terms of accuracy, parameters, and other key metrics. It is more suitable for instances involving autonomous driving when it comes to target detection.

In recent years, microservices have been very widely used as a new application development technology in edge computing, IoT, and cloud computing. Application containerization technology is one of its core technologies, which allows multiple containers to be deployed within the same physical node. Then a single physical node could provide different services to user. How to rationally deploy containers on a cluster of physical nodes is one of the main research directions nowadays. Although a number of researchers have modeled the microservice container scheduling problem and proposed effective solutions, there are still shortcomings, for example, the slow speed of finding the optimal solution and the tendency of the algorithm to fall into local optimality. This paper propose a Particle Swarm - Grey Wolf Cooperation Algorithm based on Microservice Container Scheduling Problem (PS-GWCA) by using particle swarm optimization algorithm (PSO) and grey wolf algorithm (GWO) in a multi-core parallel way, which enables the two algorithms to complement each other in the whole search process through the information exchange between populations. In the early of the search stage, the GWO can use its global search capability to guide the PSO to jump out of the local optimum to avoid premature convergence, and in the late of the search stage, the PSO can enhance the search capability of the GWO on the pareto optimal frontier. The experimental results show that compared with the other three algorithms, the algorithm optimizes 18.07% in network transmission cost, 14.67% in local load balancing, 20.66% in global load balancing, and 7.5% in search speed, and 5.69% in service reliability.

Text Classification is an important research area in natural language processing (NLP) that has received a considerable amount of scholarly attention in recent years. However, real Chinese online news is characterized by long text, a large amount of information and complex structure, which also reduces the accuracy of Chinese long text classification as a result. To improve the accuracy of long text classification of Chinese news, we propose a BERT-based local feature convolutional network (LFCN) model including four novel modules. First, to address the limitation of Bidirectional Encoder Representations from Transformers (BERT) on the length of the max input sequence, we propose a named Dynamic LEAD-n (DLn) method to extract short texts within the long text based on the traditional LEAD digest algorithm. In Text-Text Encoder (TTE) module, we use BERT pretrained language model to complete the sentence-level feature vector representation of a news text and to capture global features by using the attention mechanism to identify correlated words in text. After that, we propose a CNN-based local feature convolution (LFC) module to capture local features in text, such as key phrases. Finally, the feature vectors generated by the different operations over several different periods are fused and used to predict the category of a news text. Experimental results show that the new method further improves the accuracy of long text classification of Chinese news.