Task Assignment With Worker Churn Prediction In Spatial Crowdsourcing

With the development of science and technology,wireless networks,mobile network devices,and global positioning systems are becoming increasingly popular,human can act like moving sensors,participating in location-based tasks.Thus,the new research concept of Spatial Crowdsourcing(SC)was born.Task assignment is an important research direction in the field of SC.In SC,due to the randomness and space-time constraints of workers and tasks,there are a series of challenges in task assignment: First,an important optimization goal of task assignment is to maximize the global worker rewards,but due to the complexity of the scene,there is no polynomial complexity solution to this problem;secondly,in practice,task requesters or workers will give up after waiting for a long time,so it is required that the assignment algorithm has a high efficiency;at the same time,a SC platform needs a sufficient number of online workers,otherwise the online tasks will be unresponsive for a long time,but in practice,workers may reduce the online frequency or leave the platform due to various factors such as rewards,so it is necessary to predict the churn of workers and make biased assignment.To solve the problems as stated,the problem of task assignment based on worker churn prediction in the field of SC is proposed and studied for the first time,and a twophase framework is proposed to solve the problem,which includes the worker churn prediction phase and the task assignment phase.In the worker churn prediction phase,the Latent Feeling Capture model is used to predict a worker’s idle time interval and compare it with a given time threshold to predict workers’ churn.In the task assignment phase,a Greedy-based assignment algorithm and a KM-based assignment algorithm are proposed to assign tasks.Specifically,the Greedy-based algorithm assigns tasks to the workers who are most prone to churn,and the KM-based algorithm is to find the maximum weight matching on the bipartite graph composed of workers and tasks.In this bipartite graph,the priorities of workers who are more likely to be churn are higher.Experiments show that the proposed methods can predict and assign tasks to workers who are prone to churn.