基于声誉的协同众包激励机制研究  

Research on collaborative crowdsourcing incentive mechanism based on reputation

作　　者：刘辉 章闯 李晓皖 LIU Hui;ZHANG Chuang;LI Xiaowan(School of Computer Science and Technology,Anhui University of Technology,Ma’anshan 243032,Anhui China)

机构地区：[1]安徽工业大学计算机科学与技术学院,安徽马鞍山243032

出　　处：《华中科技大学学报（自然科学版）》2024年第9期94-100,共7页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(62306007)。

摘　　要：针对协同众包中的搭便车、不公平和不安全等问题,提出基于声誉的激励机制.该机制综合考虑成员的直接声誉和间接声誉评估其贡献,并利用区块链技术的透明性和不可篡改性存储声誉值,确保声誉数据的真实可靠.首先,采用高斯混合模型对成员任务完成的质量进行预测,并结合其历史声誉值来计算直接声誉值;然后,为了衡量每个成员在团队中的重要程度,采用聚集函数和PageRank算法.前者评估成员的局部影响力,后者衡量其全局影响力.基于成员在团队中的重要性区分度和执行任务的历史记录,分别计算成员在团队中的重要性及成员与其交互成员间的相似度对其间接声誉的影响权重,计算成员的间接声誉值.在此基础上,运用Shapley值实现收益分配的公平性,解决了成员搭便车问题.最后,基于实际数据集,将此激励机制与现有的报酬分配方法进行比较,验证其公平性和有效性.To address the problems of free-riding,unfairness and insecurity in collaborative crowdsourcing,a reputation-based incentive mechanism was proposed.The mechanism comprehensively considers the direct and indirect reputations of members to evaluate their contributions and uses the transparency and immutability of blockchain technology to store reputation values to ensure the authenticity and reliability of reputation data.Firstly,the Gaussian mixture model was used to predict the quality of members' task completion,and the direct reputation value was calculated by combining their historical reputation values,and then,in order to measure the importance of each member in the team,the aggregation function and PageRank algorithm were used.The former assessed the local influence of members,while the latter measured their global influence.Based on the importance distinction of members in the team and the history of performing tasks,the importance of members in the team and the similarity between members and their interactive members were calculated to calculate the weight of the influence of members' indirect reputation,and the indirect reputation value of members was calculated.On this basis,the Shapley value was used to realize the fairness of income distribution,which solves the problem of free riding for members.Finally,based on the actual data set,the incentive mechanism was compared with the existing reward distribution methods to verify its fairness and effectiveness.

关 键 词：协同众包 激励机制 区块链 聚集函数 高斯混合模型 SHAPLEY值 声誉 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]