An Implicit Relationship Mining Based Task Allocation Mechanism for Crowdsensing
Task allocation is a key aspect of crowdsensing,and in order to obtain high quality perception data,tasks often need to be allocated to users in a rational manner.Most existing studies simply use various optimization methods to select the best perceived users,ignoring the potential correlation between tasks and users,which leads to poor perception quality.Under this challenge,this paper designes a novel task allocation mechanism.Firstly,by analyzing the respective attributes of users and tasks,a knowledge map of the crowdsensing domain was constructed;Secondly,a link prediction method based on similarity analysis was used to fully explore the implicit relationships between users-tasks,and links were established between users-tasks with high correlation;Thirdly,in order to maximize the quality of perception data,a recruitment method that replacing individual performance with the perceived ability of user groups was proposed under the constraint of the system budget,and two feature indicators,coverage rate and reputation value,were considered compre-hensively;Finally,a group of users with the greatest sensing capability is selected based on a genetic algorithm to achieve the final task assignment.After experimental evaluation,the proposed method can effectively improve the quality of sensing data while ensuring the basic coverage of sensing tasks.
NETL
NSTL
万方数据
