To address the limitations of traditional hypernetwork node similarity methods in capturing higher-order topological structures, a method for calculating node similarity in hypernetworks based on Higher-order Distance Distribution (HDDNS) is proposed. The method generates higher-order distance distributions for nodes and employs Jensen-Shannon divergence to quantify inter-node similarity. To evaluate its effectiveness, this study conducted systematic experiments on three synthetic hypernetworks and five real-world hypernetworks, comparative analysis with five existing methods demonstrates that HDDNS exhibits superior performance in two key evaluation dimensions: mutual similarity metrics and propagation influence, the results indicate that HDDNS enables more precise and efficient computation of inter-node similarity.

To explore the risk propagation process of cluster supply chain networks under energy price shocks, based on complex network theory, numerical simulation is used to explore the risk propagation characteristics of cluster supply chain networks under energy price shocks. The results indicate that fluctuations in energy prices will have an impact on the quantity of products provided by enterprises to the next stage through production linkages; Scale free networks, random networks, and small world networks exhibit robustness against random attacks, while intentional attacks exhibit vulnerability; The fewer the first attack nodes, the slower the speed of cluster supply chain network paralysis; The probability of cluster supply chain network failure is inversely proportional to strategic resilience parameters and directly proportional to external dependency parameters.

In response to the phenomenon where some individuals on social networks verify the authenticity of rumors and then actively disseminate counter-rumors to prevent further spread, a new binary opposition information dissemination model, the CASEIR model, is proposed. The model also incorporates a transmission fatigue mechanism and analyzes through simulations the effects of different network structures, average degree, and counter-rumor initiation time on the spread of rumors and counter-rumor information. The experimental results show that the higher the average degree, the greater the impact of varying counter-rumor initiation times. Increasing the rumor reversal rate, rumor replacement rate, and information verification rate can effectively control the spread of rumors, providing a reference for the prevention and control of rumors in social networks.

Aiming at the problems that the methods for recognizing important nodes in hypernetworks ignore the influence of hyperedges on nodes and the evaluation indexes are relatively single, a method of node importance identification based on multi-indicator decision matrix is proposed. The method characterizes the local importance of node using the hyperdegree, considering the effect of hyperedges on nodes, the vector subgraph centrality is defined to portray the diffusion ability of nodes, and measures the global influence of node by reflecting its positional information through betweenness centrality, the contribution weights of each metric are determined based on entropy theory to assess the importance of the node in terms of its own influence and that of the associated hyperedges. Through the experimental validation of monotonicity, robustness, and SIR propagation model in different types of hypernetworks, the results show that the method can recognize the important nodes in the network more accurately and effectively.

To enhance the governance effectiveness of debunking texts, a quantitative analysis of their response characteristics is conducted. Based on the computational narrative network model, the semantic features of debunking texts are structured, and narrative content nodes are used to depict the audience subject and theme, while narrative logic edges are used to represent response intentions. Design a robust matching filtering algorithm to quantify the degree of counterattack of false health information in debunking texts at the semantic level. By analyzing the narrative networks of four types of debunking texts: beauty and fitness, food safety, nutrition and health, and disease prevention and control, it is found that the cohesive structure of the network is driven by emotional logic and cognitive logic. The calculation results verified the effectiveness of the algorithm and found that the response rate of debunking texts was generally polarized. The characteristics of high response debunking texts are that they match the audience and their information needs, and comprehensively use behavioral guidance, psychological identification, etc. to achieve cognitive logic construction, supplemented by authoritative, emotional and other emotional logic to enhance interactive effects, effectively intervening in the dissemination of pseudo health information.

A graph convolutional network model based on multi-scale feature fusion is proposed to address the key node identification problem in power communication networks. This model combines the inception architecture, deformable convolution, and depth-separable convolution to capture the network's node features from various scales, thereby improving the model's ability to characterize the network's complex features. In the meantime, the model's capacity to recognize the intricate details of the power communication network is enhanced by its ability to adaptively learn and assign the weights of various structural qualities during the training phase thanks to the construction of multi-channel inputs. Experiments reveal that the model outperforms standard algorithms in power communication network key node identification applications, exhibiting good performance in accuracy and robustness.

In order to accurately improve the resilience of China's electronic components industry chain, on the basis of constructing the oriented weighted network of China's electronic components industry chain using the national input-output table, the key nodes are identified and the resilience of the network is measured by considering the network connectivity contribution degree of nodes with cascading failures and the external dependence degree of nodes. The results show that for China's electronic components industry chain network, integrated circuits, sensitive components and sensors, electronic and electrical machinery special equipment, resistance capacitors and inductors, semiconductor discrete devices, electronic special materials, special glass are the key nodes with high network connectivity contribution and high external dependence.

In order to improve the management efficiency of urban public transportation system, a complex network link prediction algorithm GCNs-LP is proposed. The algorithm integrates multi-source data, designs the meta-path random walk and skip-gram model training, obtains the low-dimensional vector representation of nodes through the feature learning of neural network, and makes link prediction based on the similarity of node vector pairs. The experimental results show that the proposed algorithm achieves better prediction performance than the baseline method (CN, PA, Node2Vec, LPNMF, etc.) on the subway network data sets of Shenzhen, Shanghai and Nanjing, and the AUC value increases by at least 29%, 14.6% and 7.1%, respectively. GCNs-LP algorithm can capture the complex structure characteristics of traffic network effectively.

To investigate the evolutionary patterns of expressed and private opinions among individuals in social networks where cooperative and competitive relationships coexist, and to analyze the convergence properties of opinions, we establish a multi-leader opinion dynamics model under cooperative-competitive relationships to describe the evolution of both individuals' expressed and private opinions. By employing the randomness and aperiodicity of the influence matrix, we provide a sufficient condition for the convergence of individual expressed and private opinions, assuming the signed network satisfies specific connectivity conditions. The results demonstrate that in a balanced network structure, followers' expressed opinions are influenced by leaders' opinions, while private opinions ultimately guide the expressed opinions toward an equilibrium point. Simulation experiments further validate the effectiveness of the proposed model.

In the context of social media increasingly becoming an important platform for the spread of public opinion, social bots have malicious guidance and manipulation of online public opinion. In order to explore its guidance mechanism, this paper combines the classic Hegselmann-Krause model with the continuous action learning automaton algorithm (CALA), uses a multi-agent method to model the interaction between social bots and individual users, and captures the distribution characteristics of online public opinion through simulation. Experimental results show that social bots can effectively use the CALA algorithm to obtain network attention in different complex environments and guide the opinions of user groups to develop in the direction of predetermined goals. The study verifies the feasibility of bots in public opinion guidance.

This article focuses on the controllability of multi-agent systems with periodic sampling signals under directed star topology and switching topology. Firstly, using the Kalman rank criterion and graph theory, the decision conditions for the controllability of the multi-agent system in a directed star topology with different weights are given, indicating that the number of leaders required for the system to be controllable varies with different weights; Secondly, two controllable connection methods for star topology are proposed, and it is proved that the sampling of the system cannot be controlled when there are unreachable nodes in the topology; Finally, the PBH criterion is used to determine the sampling controllability of the multi-agent system under switching topology. It is found that the sampling controllability of multi-agent systems under switching topology is related to the joint graph topology structure.

To find new chaotic systems in electronics and circuits, a new class of four-dimensional chaotic systems based on memristor elements are proposed in this paper. We study the nonlinear dynamical behaviors of this kind of chaotic system, and finds that this kind of system has dissipative property. At the same time, we reveal the bifurcation and chaos phenomenon of this four-dimensional chaotic system from the aspects of chaotic attractors, Poincare section, the Lyapunov exponent, initial value sensitivity and bifurcation diagram by using theoretical and numerical analysis methods. Moreover, we find that the system is sensitive to the parameters and the initial values. And the system has the characteristics of multistability of chaotic flows.

To investigate the effect of noise on the discharge patterns of neurons and the synchronization stability between neurons, this paper considers the Morris-Lecar (ML) neuron model and explores its dynamical behavior by means of single-two-parameter bifurcation diagrams, Lyapunov exponent diagrams, and phase diagrams. The results show that the system not only exists rich bifurcation behaviors but also is sensitive to the initial value and exhibits multi-stable coexistence states. Next, the energy balance and synchronization patterns of the coupled system before and after the addition of noise are compared by coupling the ML neurons through a magnetic flux-controlled memristor. It is found that the coupled system can achieve energy balance and complete synchronization regardless of whether noise is added. These results provide new guidelines for identifying synaptic adaptability.

A signal bounded iterative learning control algorithm is designed to address the input saturation constraint and convergence time interval setting problem of second-order nonlinear systems. Using inequality transformation to separate unknown parameter functions in the system, establish a limited amplitude iterative learning algorithm to estimate unknown parameter functions; By means of the predefined time convergence conversion function, the arbitrary initial value sliding mode surface is converted into a new variable with zero initial value, and a variable gain limiting iterative learning controller is constructed. Strict theoretical analysis proves that the sliding mode surface converges to zero after finite iterative learning and all signals of the closed-loop system are uniformly bounded,ensuring the trajectory tracking error converges within the preset time. The numerical simulation of arbitrary initial value robot trajectory tracking control verifies the effectiveness of the proposed method and the excellent characteristics of the convergence time interval that can be set according to engineering requirements.

The personal characteristics of top executives and corporate innovation have become a key research focus in corporate finance. While existing studies primarily examine causal relationships, few explore the predictive ability, mechanisms, and patterns of executive characteristics on innovation levels. This study employs the Stochastic Gradient Boosting Tree (SGBT) algorithm to analyze the predictive power of executive characteristics. Results show that executive characteristics, particularly compensation incentives, tenure, and overconfidence, strongly predict corporate innovation levels, with nonlinear (interval) effects. These findings highlight the importance of executive characteristics in forecasting innovation.

In order to reduce unnecessary book handling and readers' waiting time, this paper takes the unmanned bookstore as the research object, sets the coordinates of the centre point of the operation area as the decision variables, and constructs an unmanned bookstore facility layout optimization model with the goal of the shortest book handling time and the largest comprehensive relationship of the operation area under the constraints of the safety distance of the operation unit, the relation between the operation area area and the overall area, and the limitation of the access gate passageway, etc, and uses the SLP method as the initial solution for the layout scheme. The layout scheme obtained is used as the initial solution of the genetic algorithm, and a simulated annealing algorithm Metropolis acceptance criterion is introduced in its cross-variation stage to propose an improved genetic algorithm for solving. After taking an unmanned bookstore as an example, comparing the optimal value of book handling time, the value of the integrated relationship between operation areas and the optimal value of the objective function of the layout scheme obtained by the genetic algorithm, SLP+genetic method, genetic simulated annealing algorithm, and improved genetic algorithm, the results show that the improved genetic algorithm is the most effective, and the layout of the facilities obtained by the scheme is more scientific and reasonable, which can effectively save the time of completing readers' orders, and enhance the customers' book buying and reading experience. The results show that the improved genetic algorithm is optimal and the proposed facility layout scheme is more scientific and reasonable.

To address the shortcomings in the field of athlete movement recognition, such as high cost of bone data annotation, unreasonable frame selection, and insufficient accuracy of recognition methods, this paper proposes an athlete movement recognition method that combines dynamic frame selection and temporal motion enhancement. Firstly, based on video RGB data, a mechanism combining global and local dynamic frame selection is proposed, and dynamic programming and adaptive beam pruning are used to optimize frame selection to ensure global consistency and local diversity. Then, a plug-and-play time and motion enhancement module is proposed, which can be arbitrarily embedded into 2D convolutional neural networks to learn video action features. Finally, the rationality and superiority of the proposed recognition method are verified on two datasets. On the HMDB51 dataset, the performance is improved by 3.16% compared to the suboptimal performance, and on the PV dataset, it is improved by 2.41% compared to the suboptimal performance. The experiments show that the combination of the global and local dynamic frame selection mechanism in the proposed method is reasonable and can balance global consistency and local diversity. At the same time, the plug-and-play time and motion enhancement module can effectively learn the spatiotemporal relationship of actions in the video and has certain superiority.

In view of the current situation of difficult data acquisition, lagging forecasting and insufficient periodicity and continuity of data mining in hotel occupancy prediction, a hybrid deep learning method of GRU-Transformer considering multi-time scale features is proposed. First, model the occupancy rate data feature matrix of year, month and day cycle time segments, and construct the input data of the model. Then, a parallel computing structure combining Attention-GRU and Conv-Transformer modules is constructed to excavate the periodic and continuity characteristics of the data and output the predicted value after integrating the characteristics. Finally, the historical occupancy data of two hotels are optimized and the ablation and comparison experiments are carried out. Experiments show that the algorithm proposed in this paper can greatly improve the prediction accuracy, and can fully meet the time requirement of timeliness. It can be used for real-time prediction of hotel occupancy data, help hotel managers timely adjust business strategy and improve hotel competitiveness.