In order to adapt to the development of dynamic network data, the detection, tracking and prediction of the community structure in dynamic networks have been a crucial research problem at present. This research reviewed the literatures on community discovery and community evolution in dynamic networks at home and abroad. This research summarized the community discovery algorithm of dynamic network, clarified the definitions of community evolution events, and sorted out the application scenarios of community evolution algorithm. Through literature review, it is believed that future dynamic community research should focus on algorithm optimization on large data sets, data mining in multiple contexts, and applicability in multiple scenarios.

As a form of public opinion, view and attitude, opinion widely exists in people′s life. It is important to clarify the evolution mechanism of opinion, explicit the existing research progress, and promote the rational governance of public opinion. In view of the lack of relevant introduction of binary opinion dynamics and the separation of the relationship between binary and group opinion dynamics, this paper summarized the research status of opinion dynamics at home and abroad. Firstly, it introduces the binary opinion dynamics model from the perspective of research method and interaction characteristic; Secondly, it combed the research results of group opinion dynamics from the perspective of individual characteristic, behavior characteristic, opinion characteristic, external environment and perspective dynamics. Finally, based on the existing research, it clarifies the problems, the mechanism of opinion evolution from the empirical perspective, the strengthening of opinions and the reduction of disputes, and the relationship between the evolution of views and group decision-making, that need to be solved in the future.

In order to study the impact of stop-loss trading on traders′ behavior and asset prices in the market, this paper constructs a multi-agent market model with stop-loss strategy based on the method of agent-based computational finance. The model simulation results show that when the stop-loss threshold is touched in the market, it is easy to trigger continuous stop-loss trading, resulting in a behavioral cascade between traders. This kind of transaction cascading leads to an increase in the convergence of trader behavior, an imbalance between sell orders and buy orders in the market, an abnormal collapse in market prices and a liquidity black hole.

Aiming at the complex dynamics problem in the evolution process of public opinion, this paper proposes a dynamic evolution model of public opinion under the vision of hyper-network. The model includes four processes: nodes addition, reconnection link, hyperedges addition and nodes aging, where node represents keywords and hyperedge represents topic of keywords. Secondly, by using the evolution mechanism of non-uniform network, a detailed theoretical analysis is carried out on the deviation of the dynamic evolution model, the deviation of the attention of nodes itself and the influence between nodes. The analysis results show that the deviation of nodes fully conforms to the power law distribution.Finally,the influence of different parameters on the variation of the node hyperdegree distribution is analyzed by MATLAB simulation, and the result is further verified to be power law distribution.

From a new perspective of opinion dynamics, this paper first clarifies the conceptual framework of acculturation of involuntary resettlers, and then constructs a cultural evolution calculation model of “social psychological modeling+complex network analysis+multi-agent system simulation”. Finally, the influencing factors and mechanism of acculturation of resettlers are quantitatively analyzed through simulation experiments. The main findings are: the cultural tolerance and susceptibility in the identity of resettlers accelerate the process of acculturation; expanding the social scope of resettlers in the resettlement areas and the guidance of elite resettlers can help them to interact with the indigenous people; policy satisfaction is more important than the creation of cultural environment, which can change the rate of cultural concept replacement of resettlers by adjusting the way of cultural shock. It also affects the stability and polarization of the process of acculturation.

Not only the development process will be seriously affected, but also it will cause a serious cascading effects when the core developers in open source software turnover. From the dynamic perspective to explore the impact by core developer turnover, and taking effective measures to protect them will promote emergence of innovation. From the perspective of complex network, taking AngularJS as an example to identify the core developers by the method of multi-attribute decision. Then we build the load capacity model to study the cascading failures by core developers turnover. The study found that the close collaboration between core developers and the core developers who occupy an important position in the network turnover will lead to seriously cascading failures; What's more, developers with larger initial workloads turnover will cause more serious cascading failure phenomenon by secondary propagation.

The structure of the convolutional neural network will also affect its performance. The design of the convolutional neural network relies more on experience and powerful computing power. How to design a neural network with better performance lacks effective theoretical support. In order to solve this problem, based on the analysis of the complexity of the typical convolutional neural network topology, in order to quickly realize the convolutional neural network that meets the given complexity characteristics, the generation from complex network topology to convolutional neural network is given. The algorithm, through the establishment of a series of convolutional neural networks with different topological features, uses the Cifar10 and Cifar100 data sets to analyze the relationship between the average clustering coefficient, average path length, graph density, modularity and other topological properties on the recognition effectiveness of the convolutional neural network. Experiments show that when the number of parameters of the neural network is basically equal, the average clustering coefficient will affect the performance of the convolutional neural network. The final conclusion is that in a statistical sense, a network structure with a small average clustering coefficient will have better performance, which provides a theoretical basis for further designing a better convolutional neural network.

In order to analyze the policy-driven mechanism of environmental protection industry (EPI) high-quality development, the connotation of EPI high-quality development is divided into five dimensions, and the system dynamics model of EPI high-quality development is constructed. Through simulation of four policy scenarios, the effect of different policy tools is investigated. The results show that there are significant differences in objective and timeliness between the four policy tools. Tax policy has the best incentive effect on EPI output, financial policy has long-term effect, science and technology policy (STP) has the strongest driving effect on output profit margin, and environmental regulation has the most prominent effect on pollution control. At the same time, there are strong complementarities among the four policy tools.

By using the stochastic frontier translog production function to calculate the agricultural technical efficiency of 31 provinces and cities from 2002 to 2017, the results show that the agricultural technical efficiency of each province has a certain degree of loss, and the average agricultural technical efficiency of western China is the lowest. Further combining with the national and regional financial support to agriculture panel data, by constructing panel data econometric model, the results show that the effect of financial support to agriculture on the agricultural technical efficiency has obvious regional differences. The relationship between finance support to agriculture and agricultural technical efficiency is linear in the eastern, western and northeastern regions, and inverted N-shaped in the central region.

For comments with multiple sections within sentences, some state-of-art models, such as Embedding from Language Models-Text Convolutional Neural Network and Generative Pre-trained Transformer model, cannot accurately extract the meaning and therefore result in unsatisfactory performance. To solve this problem, we utilize Bidirectional Encoder Representations from Transformers-Text Convolutional Neural Network and Generative Pre-trained Transformer model. Using the bidirectional code converter structure of BERT′s unique self-attention mechanism, we can obtain the word vector of the global feature of the sentence, then we input the word vectors into Text CNN, then using Text CNN to capture local features,finally we extract high-level features, such as semantics and contextual connection. This process solved the problem of inaccurate contextual connection of the text obtained by the model, allowing us to realize the fine-grained sentiment classification of Weibo comments with high accuracy. Meanwhile, to verify the advantages of the model, we compared it with existing models. The test results on the simplifyweibo_4_moods dataset show that the BERT-Text CNN model has improved accuracy, recall, and F1 indicators.