Research On Quality Control In Mobile Crowdsourcing

With the popularization of mobile intelligent terminals,it is becoming easier and easier to collect large-scale perceptual data through mobile crowdsourcing.On the one hand,the selfishness of crowdsourcing workers makes them want to get the most compensation with the least effort,which leads to the low quality of the completion of crowdsourcing tasks.However,the success of the mobile crowdsourcing task depends on the actual data contribution of each crowdsourcing worker.The quality of the data provided by the crowdsourcing participants directly affects the quality and efficiency of the subsequent research,which is more typical in this human-centered network.On the other hand,due to the uneven abilities of crowdsourcing workers,the quality of submitted data varies greatly,which makes quality control in crowdsourcing a challenge.To solve the above problems,in order to remove low-quality crowdsourcing data,find out low-quality crowdsourcing workers,and select the most suitable participants to participate in the mobile crowdsourcing task.This thesis carries out the following two aspects of work:(1)By analyzing the accuracy of the answer,the type of workers submitted to the crowdsourced data is judged,and the colluding group is judged by calculating the level of community influence after judging the existence of colluding organization,and a community influence detection algorithm is proposed.On the real data set,it is verified that the algorithm proposed in this thesis can accurately detect the colluding group leaders and their organizations.After identifying the existence of spam employees,the similarity detection algorithm proposed in this thesis is used to detect spam,and the improved similarity detection algorithm is used to screen spam employees,effectively solving the problem of failure of crowdsourcing platform caused by low quality.The effectiveness of the algorithm is verified by experiments on different data sets.Two-way quality control of crowdsourced data is carried out through these two aspects,which greatly improves the availability of crowdsourced data.(2)As participants applying for crowdsourcing tasks are greatly affected by equipment performance and other factors,the mobile crowdsourcing platform is hindered in the process of collecting data that meets quality requirements.The participant selection screening method proposed in this paper includes equipment performance and service remuneration of participants into the assessment criteria for selecting workers.The overall ability of the worker is calculated to select the most suitable participant for the perceptual task.The participant selection method is compared with the other three participant selection methods on different measurement indicators,and the comparison results show that the proposed method can increase the data quality of crowdsourcing and reduce the data overhead.