基于多Agent技术的异质社会网络群组形成方法研究

发布时间：2018-08-07 16:03

【摘要】：群组形成是目前社会网络领域的研究热点之一,可以使得若干社会个体组成团队获得更好的任务执行性能。为了成功完成任务,群组成员不仅需要拥有相关的专业技能,而且互相之间能够以团队的方式高效的合作。因此,建立一组专业的、合作高效的团队是群组形成研究中非常重要的问题。随着在线社交软件的广泛使用,社会网络给专业、合作高效的群组形成带来很大的机遇。方面,随着智能手机以及iPad等移动设备的普及,社交网站上每时每刻都会有成千上万的活跃用户,这些具有相关专业技能的用户可以帮助任务请求者建立专业的个体群组。另一方面,社会网络环境下用户之间积极的社交关系,譬如朋友之间的信任关系以及同事之间的项目合作关系,可以当作群组成员之间能否高效合作的有效指标。受上述两方面优势的启发,本文研究社会网络环境下的群组形成问题。然而,由于社会网络的开放性,自治性以及异质性特征,社会网络群组形成面临着一些新的问题。一方面,从社会个体异质性角度讲,(1)社会个体的行为是异质的,一些个体可能提供虚假的私人信息,譬如个体可能夸大自身的专业技能,社会合作关系以及工作成本来提高自身的工作报酬；(2)社会个体的目标是异质的,其中任务请求者希望建立低工作成本的群组,然而个体选择加入哪个群组以及提供哪些技能给群组任务则希望最大化自身工作报酬。(3)个体间合作效率是异质的并且是动态变化的,互相连接的不同个体之间合作效率有高低之分,并且这种合作效率在群组形成过程中是不断动态演化的；另一方面,从网络结构异质性角度讲,(4)社会网络具有社区结构特征并且社会个体的社区属性是异质的,从社会网络的中观结构角度讲,社会网络由多个具有社区结构的子网络组成,其中社区内部个体关系亲密并且合作效率高,而社区之间个体关系疏远并且合作效率低,而属于不同社区的社会个体呈现出异质的社区属性。为了解决由社会个体和网络结构异质性给社会网络群组形成带来的诸多新问题,包括(1)如何保证社会个体诚实的提供他们的私人信息,(2)如何满足社会个体之间冲突的目标,(3)如何在具有异质合作效率以及合作效率动态变化的社会网络环境下保证群组的鲁棒性,(4)如何利用社会网络的中社区结构特性提高群组效益,本文利用多Agent技术展开相应的社会网络群组形成方法研究工作,其主要贡献可以归纳如下：1) 考虑到社会个体的行为异质性,论文提出一种基于多Agent博弈论技术的激励机制。该机制通过支付与其私人信息无关的报酬来鼓励个体诚实的提供私人信息。理论分析与实验结果表明该机制保证每个个体在诚实提供其私人信息的状态下能够最大化自身收益,而且该机制能够适用于大规模的社会网络群组形成应用。2) 考虑到社会个体目标的异质性,论文提出一种基于多Agent协商技术的群组形成模型。该模型分别为不同类型的社会个体设计不同的协商策略来最大化他们各自的收益。这些策略包括任务请求者雇佣哪个个体,支付给个体的报酬,个体加入哪个群组以及提供哪些技能给群组任务。理论分析和实验结果表明,相比于传统的群组形成模型,论文提出的基于协商机制的群组形成模型能够在较短的时间开销内形成更经济的、专业的、合作高效的群组。3)考虑到个体之间合作效率的异质性以及动态变化性,论文提出一种基于移动Agent技术的分布式群组形成方法。该方法为每个任务技能分配一个移动Agent来帮助这些技能找到合适的个体群组。同时,这些移动Agent可以通过相互合作形成联盟来达到自适应网络结构的动态变化。理论分析表明,论文提出的移动Agent方法能够在多项式时间内收敛到稳定状态,并且该稳定解保证高效的群组合作性以及群组的负载均衡性。实验结果表明,相比于传统的面向静态网络结构的群组形成方法,论文提出的方法不仅能够找到专业的、合作高效的、负载均衡的群组,同时能够较好的适应动态的社会网络环境。4)考虑到社会网络的社区化特征以及个体社区属性的异质性,论文提出一种意识到社区结构的社会网络群组形成模型。该模型允许个体只能与同社区的个体合作形成群组。为了解决该社区化群组形成问题,论文提出一种启发式群组形成算法,该算法充分利用重叠社区个体的技能资源,进而优化整个系统的群组配置。理论分析证明相比于最优算法,该启发式群组形成算法在社会效益指标上具有较高的近似度,并且提出了保证该算法最大化社会效益的充分条件。实验结果表明,相比于传统的基于全局网络和邻居节点群组形成模型,论文提出的社区化群组形成模型不仅能够提高群组合作效率,而且能够提高社会效益。
[Abstract]:Group formation is one of the hotspots in the current social network field, which allows a number of social individuals to achieve better performance performance. In order to accomplish the task successfully, group members not only need to have relevant professional skills, but also cooperate efficiently with each other in a team way. Therefore, a group of specialties are established. With the widespread use of online social software, social networks have brought great opportunities to the formation of professional, cooperative and efficient groups. With the popularity of mobile devices such as smartphones and iPad, thousands of active sites are active every time. Users, those with relevant professional skills, can help task requesters to build professional individual groups. On the other hand, active social relationships among users, such as trust relationships among friends and project cooperation among colleagues in the social network environment, can be used as effective cooperation among group members. Inspired by the above two advantages, this paper studies the formation of groups in the social network environment. However, social network groups are faced with some new problems due to the openness, autonomy and heterogeneity of social networks. On the one hand, from the perspective of the individual quality of society, (1) the behavior of the individual is heterogeneous, Some individuals may provide false private information, such as individuals may exaggerate their professional skills, social cooperation and work costs to improve their own work reward; (2) the objective of the individual is heterogeneous, in which the task requestor wants to establish a group of low working groups, but the individual chooses which group to join and What skills to provide to group tasks hope to maximize their own work reward. (3) the efficiency of cooperation between individuals is heterogeneous and dynamic, and the cooperation efficiency of the interconnected individuals is high and low, and the cooperation efficiency is not dynamic in the formation of group formation; on the other hand, the network structure is different. Qualitative angle, (4) the social network has the characteristics of community structure and the community attributes of the individual are heterogeneous. From the middle view of the social network, the social network is composed of several subnetworks with community structure, in which the individual relations in the community are close and the cooperation efficiency is high, and the individual relations between the communities are distant and close. In order to solve the new problems brought by the social individuals and network structure heterogeneity to the social network group, it includes (1) how to guarantee the personal information of the social individuals, and (2) how to meet the conflict between individuals. Standard, (3) how to ensure the robustness of groups in a social network environment with heterogeneous cooperative efficiency and dynamic cooperation efficiency. (4) how to improve the group efficiency by using the community structure characteristics of the social network. This paper uses the multi Agent technology to develop the corresponding social network group formation method research work, its main contribution can be returned. As follows: 1) taking into account the behavioral heterogeneity of individuals, the paper proposes an incentive mechanism based on the multi Agent game theory. The mechanism encourages individuals to provide private information honestly by paying their private information. The theoretical analysis and experimental results show that the mechanism ensures that each individual is honest with its personal information. In the state of information, it can maximize its own income, and the mechanism can be applied to large-scale social network groups to form the application of.2). Considering the heterogeneity of social individual goals, a group formation model based on multi Agent negotiation technology is proposed in this paper. The model is designed for different types of social individuals with different negotiation strategies. These strategies include which individual is hired by the task requester, the remuneration to be paid to the individual, which group is added to the individual and what skills to be provided to the group task. The theoretical analysis and experimental results show that the group formation based on the negotiation mechanism is proposed in comparison with the traditional group formation model. The model can form a more economical, professional, cooperative and efficient group.3 within a short time cost. Considering the heterogeneity and dynamic variability of cooperation efficiency among individuals, a distributed group formation method based on mobile Agent technology is proposed. This method can help each task skill to distribute a mobile Agent to help these At the same time, these mobile Agent can form alliances to achieve dynamic changes in adaptive network structure. The theoretical analysis shows that the proposed mobile Agent method can converge to the stable state in polynomial time, and the stable solution guarantees the efficient group composition and the group. The experimental results show that compared to the traditional group formation method for static network structure, the proposed method not only can find professional, cooperative and efficient, load balanced groups, but also can better adapt to the dynamic social network environment.4). This paper proposes a social network group formation model aware of community structure. This model allows individuals to cooperate with individuals in the community to form groups. In order to solve the problem of community-based group formation, a heuristic group formation algorithm is proposed in this paper. The algorithm makes full use of overlapping community individuals. The theoretical analysis shows that the heuristic group formation algorithm has a higher approximate degree on the social benefit index compared with the optimal algorithm, and proposes a sufficient condition to ensure the maximum social benefit of the algorithm. The experimental results show that compared to the traditional global network and the traditional network, the experimental results show that the algorithm is more efficient. Neighbor node group formation model, the community group formation model proposed in this paper can not only improve the efficiency of group cooperation, but also improve social benefits.
【学位授予单位】：东南大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09;TP18

【相似文献】