In order to solve the problem of insufficient consideration of social class in the existing opinion dynamics studies, an opinion dynamics model based on the Hegselmann and Krause (HK) model is proposed by combining the topological properties of nodes in the network with the social structure on the basis of the heterogeneity of individual influences. In three types of classic complex network and an empirical network, three social structures are constructed with the social class division method adopted from the field of sociology according to capital. Simulation experiments demonstrate that the social structure has a significant impact on opinion evolution. Adjustment of the social structure through social policies is conducive to the consensus of opinions. To study the inconsistency of opinion evolution in the same type of social structure in the actual society, different initial opinion distributions of upper-class individuals are set up in the type-B pyramidal and olive-shaped social structures. It is inferred that the polarization and unification of upper-class individuals’ opinions will influence the polarization and unification of public opinion correspondingly.

The alternation of different types of edges (direction changed) and the addition of edges between different nodes in complex networks will have different effect on the controllability of the network. In order to better understand the influence of altering different edges and adding different edges on network controllability in directed networks, this paper proposes a classification method of edges. According to the node category and matching relationship, the directed edges are divided into twelve types, and the algorithm of identification is given. Based on the classification, the change law of network controllability (the number of driving nodes) is given when the edges of networks are altered and added. Through the simulation experiment of the model networks and the actual networks, the proportion of different types of edges is analyzed. When edges are altered and added, the changes in the number of driven nodes are analyzed in the model networks and the actual networks. The correctness of the theorems of this article are verified.

The cost of regular nucleic acid testing for all staff is high, resulting in long intervals and low efficiency. This paper presents a sampling strategy based on social network information, integrates online social network information with the spread of COVID-19, and establishes a novel coronavirus spread model with a sampling strategy. On this basis, this paper studies the impact of three strategies on the spread of new coronary pneumonia, namely random sampling strategy, acquaintance monitoring strategy, and large node target sampling monitoring strategy. The study found that the random monitoring strategy is better than the random monitoring strategy in the three indicators of peak amplitude, peak time, and early warning in the specific situation of the campus environment, reducing the number of days of new coronavirus transmission and the number of infected people, and more. To control the epidemic earlier and faster, the early warning effect is from high to low: target monitoring strategy for large nodes, acquaintance monitoring strategy, and random monitoring strategy.

Community partition is an important part of network research. Community partition of patients with Rheumatoid Arthritis based on medical data can effectively improve the accuracy of clinical medical decision-making. Considering that there may be problems of heterogeneity and correlation of patients′ attributes in the process of community partition, this paper firstly classifies patients based on composite rough sets theory to effectively deal with heterogeneous attributes. Secondly, the rough sets theory and the louvain algorithm are combined to build a community partition model of patients with heterogeneous attributes. By using the clinical real dataset and the classical network dataset, it is verified that the proposed model can obtain the community structure with large module value, and the realize the division of patients with different disease activity levels into different communities, so as to improve the effectiveness and accuracy of the assessment of patients′ disease activity level.

To explore the evolution of inventor influence, this paper investigates the node influence model in a multi-layer graph sequential patent citation network. Divide network layers and construct the connections between layers based on the continuity of node influence and the attractiveness of high-influence nodes. After obtaining the time series evolution data of inventor influence, the distribution and evolution law of inventor influence is explored by using piecewise fitting method. An empirical analysis of patent data in the field of ‘Molecular Biology and Microbiology’ shows that the quality and quantity of patents determine the level of influence of inventors. With high-influence inventors continuing to receive attention, most medium-influence and low-influence inventors gradually are marginalized.

In the post-truth era, the intensified antagonism between different groups and the decline of media credibility could lead to the prominent effect of hostile media effect. To reveal the influence of hostile media effect on the public opinion formation, we explore the interaction mechanism between opinion groups and mass media and propose an opinion dynamics model that considers media hostility and opinion leaders. We finally conduct simulation experiments on the model. The simulation results and analysis show that the real opinion of media, the probability of communication between groups and media, and the proportion of extreme opinion leaders can mediate the degree of group polarization. The results reveal quantitatively that the existence of hostile media effect has an important influence on the formation of public opinion, especially on the intensification of group polarization.

Considering that in the process of microblog information dissemination, the opinion of each network user is influenced by the opinions of the previous network user. Therefore, we propose a Bayesian opinion evolution model based on weibo data mining. With “dynamic zero policy is the general policy of China′s fight against the epidemic” as the key word, Python is used to crawl the Weibo comment data. After data preprocessing and word segmentation, the Bayesian opinion evolution model is empirically analyzed. Empirical analysis shows that the timely guidance of public opinion by official media plays an important role in sentiment evolution tendency.

The signing of the Regional Comprehensive Economic Partnership Agreement (RCEP) is of great economic strategic significance. To provide reference for airways to develop routes within the scope of RCEP agreement, this paper uses the theory and method of multi-layer complex network to establish an airline network model based on airways, explores the characteristics of RCEP international route network from multiple angles and the contribution of each airways in the airline network. The results show that RCEP international airline network is a scale-free network with multiple cores. All-service airways have improved the air transport efficiency among airports in RCEP international airline network, while low-cost airways are the main force in expanding the coverage of RCEP international airline network.

Aiming at the difficulty of balancing the accuracy and complexity of the sign prediction algorithm, this paper effectively integrates the law of social development and the local characteristics of the network, and proposes a sign prediction algorithm based on structural balance theory and status theory to calculate the similarity of nodes. In order to better combine the contribution of the above two theories to the similarity score of the two nodes, this paper uses the regulator to sum the similarity score based on the two theories according to the weight of the regulator, and the positive or negative of the similarity score is the result predicted by the edge symbol. Finally, the algorithm is tested on several different data sets and compared with the classical CN algorithm and PSNBS algorithm in two aspects of prediction accuracy and algorithm complexity. The proposed algorithm is very close to the classical algorithm in terms of prediction accuracy, but in terms of time complexity, it is an order of magnitude lower than the classical algorithm.It is obviously better than classical algorithm.

In order to solve the increasingly serious congestion of expressways, the traffic operation state and its evolution law are discussed, and the traffic state discrimination standard is determined. The traffic simulation platform VISSIM software is used for simulation analysis, and the spectral clustering analysis algorithm is used to accurately extract the abrupt boundary of traffic state. Combined with the cusp catastrophe theory, the traffic flow state at different times and locations is studied and analyzed. The results show that there is an obvious catastrophe phenomenon when the traffic flow transforms between the two states of free flow and congestion flow.

This paper comprehensively considers the impacts of abatement investment’s lag effect and consumer’s low carbon reference effect on the market demand and brand goodwill, constructs a dual-channel low carbon supply chain system consisting of a manufacturer and a retailer, and adopts optimal control theory to compare and analyze the members’ operation strategies, product low-carbon level, brand goodwill, and overall profit under the decentralized, centralized, and cost sharing decision-making modes. The study finds that reference effect always mobilizes the manufacturer′s enthusiasm for low-carbon production, improves the steady-state value of low-carbon level of product and brand goodwill, and discourages companies from advertising investment; It is meaningful for members of the low carbon supply chain to adopt the centralized decision-making mode only when the delay time is below a certain threshold; The delay time raises consumer’s psychological expectation of the low-carbon level of product, reference effect intensifies the impact of delay time on the overall profit of supply chain under different decision mechanisms.

In order to analyze the influence of individual behavioral characteristics and decision making preferences on the evolution of water resources conflicts, the authors introduced prospect theory into the evolution game of transboundary river water resources conflict analysis, applied the prospect value function to construct the perceived benefit matrix of the water resource conflicts between upstream and downstream regions, used evolutionary games to analyze the evolutionary paths and results of decision-making behaviors of upstream and downstream regions, and constructed a system dynamics model for simulation analysis. The results show that: upstream and downstream regions need to meet the complex game conditions for selection of the cooperation strategy set, and when the perceived cost of active water release from upstream, the perceived income of passive water release from upstream, the perceived cost of active pursuit increase from the downstream, and the perceived costs of water conflict risks decrease, the probability of choosing the cooperation strategy set will be reduced, and water resources conflict will occur.

For the robot formation tracking problem of unknown target, a robot motion control model is established, and a target tracking and ring-around control strategy based on Reinforcement Learning(RL) is proposed to solve the problem. Driven by RL, the robot explore the location of the target point and initiate tracking. The robot tracking strategy is optimized in real time using the ring-around formation motion model to achieve dynamic tracking and ring-around control of the fleeing target point. A multi-robot motion control environment is established, and the experiments indicate that the combined RL can accelerate the multi-robot formation adjustment time and prove the efficiency of the multi-robot ring-around formation control strategy.

In order to improve the utilization of hardware resources of driverless vehicle perception system, a multi-task driverless vehicle perception algorithm based on feature fusion is constructed. The improved CSPDarknet53 is used as the backbone network of the model, and multi-scale features are extracted and fused by constructing feature fusion network and feature fusion module. The detection of 7 common road objects and pixel-level segmentation of the driving area are taken as examples. Multi-task DaSNet (Detection and Segmentation Net) is designed for training and testing. In order to compare model performance, BDD100K data set is used to train YOLOv5s, Faster R-CNN and U-NET models, and comparative analysis is made on mAP, Dice coefficient and detection speed and other performance indicators. The results showed that DaSNet multi-task model′s mAP value is 0.5% and 4.2% higher than YOLOv5s and Faster RCNN, respectively, and the detection speed of 121FPS can be achieved on RTX2080Ti GPU. Compared with U-NET network, Dice value of segmentation in priority and non-priority drivable are 4.4% and 6.8% higher, showing an obvious improvement.