Improvement And Application Of Crowdsourcing Task Assignment Algorithm

Crowdsourcing markets have become very popular among researchers as a way to collect data inexpensively using large numbers of Internet workers.However,crowdsourcing is weak for managing workers,that is to say,how to assign diverse tasks to workers with different unknown skills is still an urgent problem to be solved.Dual Task Assigner(DTA)is a new task assignment algorithm which solves the problem of assigning diverse tasks to workers with unknown skills in crowdsourcing environment.The DTA algorithm,however,can not well identify dishonest workers and can not adequately use crowdsourcing workers at high skill levels.In the thesis,we will improve the DTA algorithm to make it more efficient at task assignment.In the thesis,a worker selection algorithm is proposed firstly to select the honest crowdsourcing workers to enter the subsequent task assignments.Compared with the Automatic Retrieval System(ARS)of the crowdsourcing platform,it is found that the worker selection algorithm we proposed can effectively select the honest workers and reduce the impact of dishonest workers on subsequent task assignments.Secondly,based on the iterative results of the worker selection algorithm,the skill level values of workers are calculated.And the baseline weight values of each task are calculated using the Artificial Bee Colony(ABC)algorithm.Then,as for the problem of single assignment strategy of DTA algorithm,the thesis adds another two assignment strategies.One is assigning more tasks to the high skill level crowdsourcing workers,the other is not to assign any task to the crowdsourcing workers with insufficient skill level.Thus,the high skill level crowdsourcing workers can be used adequately,and ensure that the final outcome of the task has a higher quality.Finally,through experimental analysis based on the real data collected in xTurk,we find the improved DTA algorithm performs better than the DTA algorithm and random assignment algorithm.At the end of this paper,we use Python and Django framework to build a simple Crowdsourcing website,and the allocation strategy of the improved DTA algorithm is implanted to the site to achieve a reasonable distribution of tasks,users deepen understanding of the task allocation process of the improved DTA algorithm by the simple interaction process.