Research On Three-dimensional Assignment Mechanisms Of Crowdsourcing Tasks With Preference

With the continuous development of Internet technology,crowdsourcing,as a new form of production organization,has received widespread attention and developed rapidly.Crowdsourcing distributes and executes tasks through the Internet,and can effectively reduce the task execution costs and improve the execution efficiency.The proper task matching mechanism can increase the value generated after the tasks are completed.Therefore,the matching mechanism of tasks is particularly important.The incentive mechanism can effectively stimulate more workers to participate in crowdsourcing activities and improve work efficiency.In the crowdsourcing systems,task requesters usually have preference over workers due to various factors such as professional level,effort,geographic location.On the other hand,location-related crowdsourcing services require task requesters and workers to go to the locations independently.These have brought new challenges in the assignment of tasks.For the novel crowdsourcing scenario with preferences and independent locations,this thesis proposes a crowdsourcing distribution model that considers preferences and independent locations.The crowdsourcing platform first releases worker information to all task requesters.Each requester submits a task request to the platform,including task information and a set of preferences over workers.The platform combines the local workplace information for task assignment.The workers arrive at the designated locations to complete the assigned tasks and upload the data to the platform.Finally,the platform provides the data to the requesters.For this model,we design a crowdsourcing task allocation mechanism that considers preferences and independent locations.The goal is to maximize the total value of completed tasks.We consider two models: In the first model,we assume that the value of completing every task is the same.Using the greedy rule and local search algorithm,we perform path decomposition on the initial solution obtained by the greedy algorithm to search for an improved final solution;The second model relaxes the value constraint,assuming the value of completing every task is different.We combine the greedy rule and the local search algorithm by searching for an improved set with a return factor greater than 1 in the neighborhood over the initial solution.After theoretical analysis and experiments,comparing with the classic greedy algorithm in terms of the total value of the task,running time and memory consumption,we show that the designed matching algorithms have good properties of computational efficiency and constant approximation ratio.