移动群智感知中基于协同排序的任务推荐方法  

Task Recommendation Method Based on Collaborative Ranking in Mobile Crowd Sensing

作　　者：王健[1] 刘嘉欣 赵国生[2] 赵中楠[1] WANG Jian;LIU Jia-xin;ZHAO Guo-sheng;ZHAO Zhong-nan(School of Computer Science and Technology,Harbin University of Science and Technology,Harbin,Heilongjiang 150080,China;College of Computer Science and Information Engineering,Harbin Normal University,Harbin,Heilongjiang,150025,China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,黑龙江哈尔滨150080 [2]哈尔滨师范大学计算机科学与信息工程学院,黑龙江哈尔滨150025

出　　处：《电子学报》2021年第10期2012-2019,共8页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61403109,No.61202458);高等学校博士学科点专项科研基金(No.20112303120007);黑龙江省自然科学基金(No.LH2020F034)。

摘　　要：针对移动群智感知中参与者积极性不高导致的数据质量低和激励成本高的问题,本文提出了一种基于混合用户模型与列表级排序学习算法相结合的协同排序任务推荐方法.根据参与者的历史行为对其进行分析,初步过滤掉一些劣质感知用户,同时利用参与者间的相似性构建混合用户模型.利用概率矩阵分解对参与者的意愿值进行预测,并根据排序学习得到一个排序模型.根据排序模型生成任务推荐列表,作为目标参与者的优选任务列表.基于真实数据集的仿真实验结果表明,本文提出的方法有效地提高了任务分配的准确率,与此同时减少了感知用户的移动距离.To solve the problem of low data quality and high incentive cost caused by the inactivity of participants in mobile crowd sensing,this paper proposes a task recommendation method.We could analyze the participants according to their historical behavior,and filter out low-quality sensing users.Meanwhile,the similarity among the participants was used to build a user-hybrid model.Then,the participants'willingness would be predicted by the probabilistic matrix factorization,and a ranking model was obtained.Finally,a task recommendation list was generated on the basis of ranking model as the preferred task list for the target participants.The simulation experiments based on the real dataset show that the proposed method in this paper can improve the accuracy of task assignment effectively and reduce the moving distance of sensing users simultaneously.

关 键 词：移动群智感知 任务推荐 协同排序 混合用户模型 参与者意愿 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]