Robust Crowdsource Data Analysis Method Based On Trust Model And Its Application

Crowdsourcing is a very popular new business model derived from the great development trend of the Internet.Enterprises distribute the tasks previously performed by employees to non-specific(usually large)public volunteers in a free and voluntary manner to complete them,and follow the open and brainstorming ideas to obtain the highest quality task results.Volunteers can get a lot of rewards after they have made their own efforts and completed their tasks.This is a brand-new labor subcontracting mode that crowdsourcing provides for software industry and service industry in the current Internet era.A large number of volunteers have benefited from the mode of work provided by crowdsourcing and received a lot of rewards.However,in this process,some volunteers did not complete the task seriously.In order to cheat commissions and maximize benefits,they often provide false data.Once the data provided by such volunteers are adopted and used,it will bring great losses to enterprises.Therefore,evaluating and screening the quality of crowdsourcing task results is a challenging task.At present,the research on crowdsourcing quality control at home and abroad is still in the initial stage.In view of the above problems,this paper studies the quality evaluation of crowdsourcing data and proposes some effective methods to obtain high-quality crowdsourcing data,mainly including the following aspects:(1)This paper studies the development of crowdsourcing and summarizes the existing methods of crowdsourcing data quality evaluation.This paper studies the Bayesian algorithm model,and summarizes and analyzes the application of Bayesian model in different situations.(2)This paper proposes a robust crowdsourcing data analysis method based on trust model.The historical behavior data of workers are taken into consideration,and the result data submitted by workers in the current task are analyzed.Finally,the historical behavior data of workers and the result data submitted in this task are combined with Bayesian algorithm model to obtain the final accuracy information of the result data provided by workers in this task.Compared with the gold standard data evaluation strategy,the results show that the workers screened by this method have higher reliability.(3)This paper proposes a dynamic replacement strategy for workers based on trust model.First of all,before the task starts,workers are preliminarily screened and a pool of candidate workers is constructed.Then,the crowdsourcing task is divided into N stages.After the completion of each stage of the task,the result data provided by the workers in this stage are analyzed through voting consistency rules.Once it is found that the accuracy of the result data submitted by the workers cannot meet the needs of the employers,the unqualified workers are marked and the unqualified workers are replaced in time,thus ensuring that the quality of the task results submitted by the workers always meets the needs of the employers during the task process.After the completion of all task stages,the result data submitted by the workers can be directly adopted and used by the employers.