Research On Low Cost Crowdsourcing Allocation Method Combining Task Attribute And Worker Profile

In recent years,with the research and development of crowdsourcing mode,more and more research projects are devoted to optimize all the process stages of crowdsourcing,improve the satisfy of users and save the cost of task providers.As one of the hot research fields,crowdsourcing task allocation has a considerable research prospect.With the increasing number of users participating in crowdsourcing,the more task providers using crowdsourcing mode to achieve their goals,the number of complex interactions between tasks and users are increasing,so a task allocation method that users like and task providers satisfied is particularly important.It can effectively complete the matching between users and tasks in the task allocation stage,which will increase the satisfaction of users to the crowdsourcing system,and effectively reduce the reward cost of task providers.In the existing crowdsourcing task assignment research,one is to assign tasks to users directly through the platform,and push them to the appropriate users according to the existing tasks,but the user's interest in the tasks may affect the quality of task completion.Some researches put forward the pull-down method,in which users actively search to display the corresponding tasks,so as to achieve task assignment.Although these models solve the problem of users' interest,however,they can't fully realize the assignment of all tasks,and some tasks may be users interested but they don't know.With the continuous development of artificial neural network,researchers leverage user profiles to assign tasks to users.But this assignment may lead to mislead assignment,and more popular tasks will get more exposure opportunities.So task providers will have to increase the reward to facilitate other tasks to be completed,resulting in higher allocation costs.Therefore,this study tries to break the bottleneck,design a scheme to ensure the user's interest distribution,and at the same time,it can effectively save the reward cost of task allocation.In the design process,we will face the problem of ensuring the satisfaction of users is not affected,and how to make most tasks get exposure opportunities,and low-cost completion of the task assignment while these tasks no user selection.Through the preliminary analysis and collection of volunteer questionnaire survey,the study found that users in crowdsourcing select tasks always according some special attributes.Different attributes in tasks have different influence on users.Users seldom measure the comprehensive results of all attributes carefully when making decisions.Some attributes that they especially like or dislike have great influence on users' decisions.According to this finding,this paper proposes the APLoc Assign crowdsourcing task allocation method.This method not only considers the coverage of user profile and task allocation,but also can effectively guarantee the user's interest in the task allocation and ensure the user's satisfaction with the crowdsourcing platform.It can also reduce the reward cost of task allocation and provide task providers with a low-cost and high-quality task completion result.After experimental verification,the APLoc Assign task allocation method can achieve accurate and effective allocation for real experimental scenarios,and also shows excellent performance on large-scale data sets.