Research On Crowd-based Knowledge Acquisition Model Based On Social Network Interaction

With the development of science and technology,social network platforms emerge in large numbers.The way people access information has become diversified.The social network platforms realize instant interaction and resource sharing based on people's social relationships,which greatly accelerates the process of information propagation.People are no longer just recipients of information,they are also the disseminators of information.Recently,people begin to pay attention to the diffusion problem on social networks to observe the dissemination process of information,opinions,rumors on social networks and explore the truth hidden behind social phenomena.In the study of diffusion,the construction of diffusion model and the analysis of influence are particularly important.The construction of diffusion models can help us explore the patterns of social activities.The analysis of influence on diffusion models can help us control the spread of diffusion guiding the influence in a favorable direction.At present,these diffusion models mainly simulate the diffusion process of infectious diseases,information,innovation,and products.As we all know,the spread of knowledge is crucial to the development of human society.Knowledge is the interpretation and refinement of data by high-level expression,and it is an increasingly common sight to disseminate knowledge based on social networks.However,the research of diffusion model fails to display the diffusion process of knowledge.Moreover,the traditional diffusion models often use nodes to represent social participants,and nodes only have the finite states(activated state,inactive state,etc.)in the process of diffusion with certain limitations.So it difficult for us to simulate the exchange of knowledge between people.Based on it,this paper proposes a Crowd-based Knowledge Acquisition model based on social network interaction,which aims to simulate the diffusion process of information and knowledge in the population.Individuals in the social network are no longer nodes,but intelligent agents with the ability to perceive the environment,process information,and judge knowledge.This paper focuses on the characteristics of the network structure and its impact on knowledge diffusion.Specifically,the model consists of multiple agents interactions through a social network structure and uses a decentralized manner to collect and process data.There,data represents information,and agents have the learn ability to internalize information into knowledge,and can improve their knowledge models through exchange.The model has two phases.Initialization phase,each agent collects its own data at the beginning to generate an initial model of knowledge,then exchange information to update the knowledge models.Interaction occurs in a social network,including information as well as knowledge.We assume that the interaction between agents is two-way.The interaction of data extends the information received by the agent,and the interaction of the knowledge model allows the agent to adopt one based on the accuracy of the models over the local data sets.Through iterations of interactions,the agents are able to arrive at a unified knowledge model eventually.We use decision trees as the knowledge model for individual agents because decision trees are natural models of association rules among data attributes.Not only they are commonly used in knowledge acquisition,but also when putting into an ensemble,a collection of randomly constructed decision trees often exhibit remarkable robustness and reliability.The research is mainly divided into the following four aspects:(1)At the micro level,we focus on the location attributes that affect the accuracy of an individual knowledge model,and explore whether the centrality of the agent can provide an advantage for knowledge acquisition.We find that the traditional diffusion model such as Independent Cascade Model does not truly reflect the impact of individual location on knowledge acquisition.Our model shows that agents with low eccentricity have an advantage in knowledge acquisition.(2)At the macro level,the paper focuses on global structural metrics such as density and clustering coefficient and study their significance in the average accuracy of the decision trees.We also compare some standard network structures such as complete,star,path,scale-free,small-world and random networks.(3)Based on Crowd-based Knowledge Acquisition model,the paper puts forward a new definition of knowledge-based influence which is measured in terms of the extent in which a set of agents affect the knowledge model of other agents through the exchange.The study finds that the influence maximization problem has a certain correlation with the knowledge influence problem,and the influence maximization algorithm can also be used to solve the knowledge maximization problem.(4)In order to verify the rationality of our model,we conducted a large number of experiments.We used 14 synthetic networks,including small-world,scale-free,random networks,and 3 standard networks,including complete,star,path networks,and 7 real-world social networks,including the trade network,marriage network,club network,dolphin network,Facebook network.At the same time,we also used 3 different supervised learning data sets.The experimental results show that our mechanism can promote the diffusion of good knowledge models,and also proves many important structural insights,which are highly consistent with real social situations.