Modeling And Simulation For Crowdsourcing Platforms From The Perspective Of Complex Adaptive Systems

Crowdsourcing is a new kind of business which utilizes the internet to distribute tasks to the online workers around the world.Because of its advantages in time and costs,crowdsourcing has been widely applied to many fields.Enterprises can reduce the production costs and raise the production efficiency.Since there is no contract between requesters and workers,the quality and schedule control have become a crux to determine whether or not a crowdsourcing project succeeds.Because a crowdsourcing system is composed of a large number of workers and the dynamics results from the behavior of individual workers and the interaction among workers,a crowdsourcing system is a complex adaptive system.On the other hand Agent-based modeling(ABM)also consists of a large amount of autonomous entities.Therefore,ABM is an ideal way to study complex systems.This thesis studies the dynamics of crowdsourcing systems from the perspective of complex adaptive systems,with ABM as the tool.We analyze the behavioral rules of the online workers through experiments,and found that the accuracy and the completion time of the tasks follow Lognormal distribution.Then a multivariate linear regression model is established to capture the relationship between the task parameters and the accuracy and completion time.In order to further improve the accuracy,we assign tasks in both the parallel and the sequential way and derive the behavioral rules of the online workers.By assigning tasks in parallel and sequential ways,the interaction among online workers is built.On the basis of the interactions,an efficient social network emerges,which is beneficial to further enhance the quality and efficiency of the tasks.We found that the accuracy and efficiency in the parallel mode is higher than those in the sequential mode,and the accuracy improves with the increase of the number of tasks.Finally,in order to obtain higher accuracy and efficiency,the Genetic Algorithm is applied to optimize the parameters of the task to figure out the optimal task posting strategy given a target accuracy and completion time.As a test-bed,all kinds of task posting strategies can be tested on the ABM,which serves as a foundation for further improvements.