移动群智感知中基于用户意愿的参与者优选方法  

Participant selection based on user willingness in mobile crowd sensing

作　　者：吴佳莹 张振宇[1] 韩俊樱 WU Jia-ying;ZHANG Zhen-yu;HAN Jun-ying(College of Information Science and Engineering,XinJiang University,Urumqi 830046,China)

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046

出　　处：《东北师大学报（自然科学版）》2021年第3期72-80,共9页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61262089).

摘　　要：移动群智感知中,在选择参与者时忽视用户意愿,可能会导致用户执行效率低甚至产生中途退出、上传虚假数据等严重损害感知数据质量的行为.针对这一问题,首先提出了一种基于用户意愿的参与者优选方法,综合考虑用户特征、任务属性、实时环境等信息,使用全连接的深度神经网络以及卷积自编码器构建用户意愿回归模型,量化评估在实时过程中用户对感知任务的执行意愿程度.其次设计平衡用户意愿与用户效用的参与者优选机制,在不同场景下同时保障感知效率与用户意愿.通过在真实轨迹数据集上进行仿真实验,验证了用户意愿回归模型的拟合度优于基线方法.通过两个感知任务案例分析验证了平衡用户意愿与用户效用的参与者优选机制的有效性及合理性.In the mobile crowd sensing,ignoring the user s willingness in the participant selection process may lead to inefficient user execution or even untrustworthy behaviors that seriously damage the perceived data quality,such as withdrawing halfway and uploading false data.Aiming at this problem,a participant selection method based on user will-ingness is proposed,which takes into account user characteristics,task attributes,real-time environment and other information,and uses a deep neural network and a convolutional auto-encoder to construct a user willingness regression model.The degree of user willingness to perform a perceived task in real-time.Secondly,a participant optimization mechanism is designed to balance user willingness and user utility,so as to ensure both perceived efficiency and user willingness in different scenarios.Multi-Agent simulation experiments on the real trajectory data set verify the effectiveness of the user willingness regression model,and its evaluation accuracy is over 90%.

关 键 词：用户意愿 深度神经网络 卷积自编码 参与者选择 移动群智感知 

分 类 号：TP393.01[自动化与计算机技术—计算机应用技术]