Task Assignment Based On Multi-objective Control For Crowdsourcing

At present,the rapid development of artificial intelligence,big data,and other fields has brought great convenience to our life.Many data need to be tagged for the field of machine learning,data cleaning,etc.Crowdsourcing technology provides a new way of publishing problems on the Internet,and settling problems through the wisdom of the masses on the Internet.In crowdsourcing technology,quality,economic cost and latency cost are three main research objectives.In the task assignment of crowdsourcing,researchers tend to focus on improving task quality and ignore the economic cost and latency cost.The existing research on multi-objective control is limited.Moreover,the existing researches depend on the specific task type,and there is no general control method.In this paper,all the three research objectives in the crowdsourcing are taken into account,and the general multi-objective control model is put forward combined with task assignment.Existing solutions for answer aggregation in crowdsourcing technology are based on bayesian decision algorithms,and task assignment is under the condition that we need to predict the task answer.An answer prediction algorithm suitable for task assignment phase is put forward in this paper based on the bayesian model;At the same time,a variety of algorithms are proposed to meet the two multi-objective control models,and the effectiveness and efficiency of the algorithm are evaluated.In crowdsourcing technology,economic cost is due to worker incentive.Existing incentive model is based on task searching model.A kind of task assignment model suitable for workers incentive is put forward and optimized.Furthermore,the causes of latency cost in task assignment are analyzed,and the way of latency cost control is proposed.Our task assignment strategy is implemented in CrowdIQ crowdsourcing platform.This paper gives an overview of CrowdIQ crowdsourcing platform,analyzes and summarizes the multi-objective control in CrowdIQ.