Table of Contents (Volume #10, Issue #3)

         Cover Page (Volume #10, Issue #3)

         Editorial (Volume #10, Issue #3)

Paper #1

Spatiotemporal Transformer-Based Video Generation Method for Multi-Camera Smart City Scenarios

Bingchao Sun

Abstract: Traditional video acquisition and generation methods are limited by sensor layout, data transmission costs, and modeling capabilities, rendering it challenging to simultaneously capture local structure and global temporal dependencies in complex, multi-scale, and irregular urban scenes, resulting in inconsistent video quality and stability. To address the problem of urban video frame reconstruction and generation under complex monitoring conditions, this study introduces a smart city video generation method that integrates spatiotemporal transformer networks and graph convolution networks. First, a spatiotemporal Transformer is used to capture spatiotemporal semantics. Then, graph convolution is introduced to model structural relationships among spatial patches within video frames, enabling effective spatial information propagation. Furthermore, gated recurrent units and multi-scale attention mechanisms are combined to enhance dynamic modeling and feature fusion capabilities. Experiments are conducted on urban monitoring video data containing three representative scenarios, including urban road traffic monitoring, public square surveillance, and environmental monitoring scenes. The performance of the proposed method is compared with convolutional neural network based video reconstruction methods and spatiotemporal transformer based models under the same experimental settings. In these scenarios, the introduced approach achieves a peak signal-to-noise ratio improvement of 2.14 dB, 3.01 dB, and 3.40 dB compared to the convolutional neural network based reconstruction method, respectively, while the structural similarity index increases by approximately 1.52% to 2.97% across different scenes. Moreover, compared to existing video modeling methods reported in the literature, the proposed method exhibits the least performance degradation under various types of noise and occlusion disturbances. The findings demonstrate that the introduced approach can achieve higher quality and more robust video generation in complex dynamic scenes, particularly in urban monitoring environments with noise interference and partial occlusion, providing a new technical path and implementation means for visual perception in smart cities.

Keywords: Graph Neural Network; Multi-Camera Video Generation; Multi-Scale Feature Fusion; Smart City; Spatiotemporal Transformer.

Download Full Text

Paper #2

Real-World Application of a Stochastic Three-Objective Optimization Model for Greening, Cost Reduction, and Time Management in a Blood Supply Chain Network: A Case Study

Xu Guo

Abstract: Managing a green Blood Supply Chain (BSC) under stochastic demand and perishability constraints presents a critical optimization challenge. This study addresses this by proposing a novel four-echelon, multi-objective Mixed Integer Programming (MIP) model designed to minimize network costs, transportation time, and environmental impacts (waste and emissions) simultaneously. Unlike previous deterministic models, this research incorporates stochastic demand utilizing a chance-constrained programming approach and explicitly models the shelf-life constraints of three distinct blood products (red blood cells, platelets, plasma). To handle the conflicting objectives, the Lp-metric method converts the multi-objective model into a single-objective formulation, enabling efficient solution via the CPLEX solver in GAMS. The model's applicability is validated through a numerical case study, followed by a comprehensive sensitivity analysis on waste ratios and demand fluctuations. Results demonstrate that the proposed framework effectively balances cost-efficiency with sustainability, reducing waste by optimizing flow allocations. Furthermore, the study discusses the computational trade-offs between exact methods and heuristic approaches for large-scale implementations. This research bridges the gap between theoretical green logistics and practical healthcare supply chain management, offering a robust decision-support tool for blood center managers to navigate uncertainties in demand and product shelf-life.

Keywords: Blood Supply Chain; Green Logistics; Lp-metric; Shelf Life; Stochastic Optimization.

Download Full Text

Paper #3

A Residual Feature-Gated Variational Autoencoder with Transfer Learning for Economic Systemic Risk Exposure Prediction

Shiqiong Pan

Abstract: Accurate prediction of systemic economic risk is important for financial supervision and policy design, yet existing econometric and machine-learning models often suffer from limited data and unstable generalization. This paper proposes a two-step learning framework for predicting systemic risk exposure using a simulated economic resilience dataset. First, a Variational Autoencoder (VAE) is trained on real observations and used to generate 100,000 synthetic samples for data augmentation. Second, a residual neural network with a feature-gating mechanism is pretrained on the synthetic data and then fine-tuned on real samples through transfer learning. Experiments show that the proposed strategy leads to large performance gains. Compared with training from scratch, the two-stage model improves R² from 0.958 to 0.998, while reducing RMSE from 0.015 to 0.003 and MAE from 0.011 to 0.003. Classical models also benefit from augmentation: for SVR, R² increases from 0.018 to 0.090 after adding synthetic data. Feature screening based on mutual information highlights growth dynamics and risk-control indicators as the most influential drivers of systemic risk. Ablation studies further show that residual connections and feature gating are both essential, with the full model achieving the lowest error levels. These results demonstrate that generative data augmentation combined with transfer learning provides an effective and general solution for economic risk prediction under limited data conditions.

Keywords: Data Augmentation; Financial Stability; Systemic Risk Prediction; Transfer Learning; Variational Autoencoder.

Download Full Text

Paper #4

Intelligent Scanning and Remediation Algorithms for Information Security Vulnerabilities in IoT Terminals

Pengcheng He and Weiju Kong

Abstract: The number of IoT terminal devices is growing rapidly. However, due to limited computing resources, diverse architectures, and difficulties in updating algorithm models, traditional vulnerability detection and fault repair mechanisms are inapplicable. Therefore, this paper constructs a security vulnerability scanning algorithm for IoT terminals. The IV-PARM model constructed in this paper combines multiple features, including data from three modalities: firmware semantics, runtime behavior, and network traffic, to build a heterogeneous graph neural network. Device vulnerability association inference, combined with risk perception, is generated based on a reinforcement learning model. An adaptive and lightweight repair strategy is used to guide the update of IoT terminals. The model framework of this paper adopts a three-level collaborative approach, which can reduce the resource overhead of the terminal for vulnerability detection while ensuring the accuracy of vulnerability detection. This paper conducts experiments on 1200 real samples and 120,000 behavior logs. The experimental results show that the IV-PARM model can achieve an accuracy of 85.7% in vulnerability detection, with a false positive rate of only 2.3% and a repair success rate as high as 91%. On the other hand, in terms of latency, the average latency of the model is less than 45 milliseconds, and the memory usage is less than 15 MB. The experimental results in this paper demonstrate that the proposed model has significant advantages in terms of effectiveness, robustness, and inference..

Keywords: Adaptive repair; Graph neural network; IoT security; Intelligent vulnerability detection; Reinforcement learning.

Download Full Text

Paper #5

Graph Clustering Research Based on Adaptive and Optimal Probabilistic Graph Learning

Lingguo Zou and Meihua Zhang

Abstract: With the continuous development of mobile internet, related data has experienced explosive growth. However, existing graph clustering methods suffer from issues such as high sensitivity to the quality of initial graph structures, the disconnect between graph construction and representation learning, and difficulty in handling graph structural noise. This research proposes a graph clustering model based on adaptive graph learning and optimal probabilistic graph learning. First, to address the limitation of static graph structures that cannot dynamically optimize based on node representations, the Adaptive Graph Clustering (AGC) model is proposed. Through a learnable adjacency matrix update mechanism, the graph structure adapts to serve the clustering objective during training. Second, to address the limitation of existing deterministic graph learning ignoring edge uncertainty, the Optimal Probabilistic Graph Clustering (OPGC) model is proposed. It models the graph structure as a Bernoulli distribution, learning edge existence probabilities through probability-weighted graph convolutions and entropy regularization, thereby explicitly quantifying structural noise. Additionally, a dual self-supervised clustering strategy is designed to simultaneously generate soft assignments from both attribute and graph spaces while minimizing Kullback-Leibler divergence from the target distribution. This unifies attribute feature learning and probabilistic graph structure learning within an end-to-end framework. The adaptive graph clustering model demonstrates outstanding performance on both structural and textual data. It achieves a maximum clustering accuracy of 80.2%, surpassing the second-best deep adaptive graph clustering by 7.8%. After removing 15% of original edges, the model's normalized mutual information decreases by only 0.044. The optimal probabilistic graph clustering model achieves a clustering accuracy of 0.852. Ablation experiments demonstrate that the adaptive graph learning module is most critical for enhancing model performance. The model converges 26.9% faster than graph attention networks and consumes 23.6% less memory. Consequently, the proposed model effectively improves graph clustering accuracy and robustness, significantly boosts computational efficiency, and simplifies complex networks.

Keywords: Adaptive graph learning; Graph clustering; Graph convolutional neural network; Optimal probability; Self-supervised learning.

Download Full Text