Research On The Benefit-optimal Task Scheduling Algorithms For Crowdsourcing Platform

With the rapid development of the Internet,interpersonal communication has become easier,and the widespread use of smart devices has facilitated the ability of ordinary users to collect and deliver large amounts of information.In this context,Crowdsourcing(CS)has attracted a lot of attention in recent years as a paradigm for efficient and distributed problem solving.Crowdsourcing is based on the Internet’s large user community,integrating the wisdom of scattered individuals or groups to effectively solve tasks that are difficult to accomplish by machines,reflecting the power of collective intelligence.Crowdsourcing platform,as a key component of the crowdsourcing system,connects crowdsourcing workers and task requesters,and involves almost all aspects of the crowdsourcing process,which is crucial to the completion of tasks in terms of quality and efficiency.Therefore,this paper,oriented to the goal of optimizing the benefits of crowdsourcing platform,investigates four task scheduling scenarios in crowdsourcing platform with crowdsourcing tasks as the throughline,which has certain research significance for promoting the improvement of service quality of the whole crowdsourcing system.Specifically,this paper firstly studies the task dissemination problem in the crowdsourcing platform and proposes using social networks as a platform for task dissemination and worker recruitment,which expands the number of workers and task coverage in the crowdsourcing platform,solves the platform cold boot problem,and improves the benefits of the platform.Secondly,this paper studies the task allocation problem in crowdsourcing platforms.Based on the real crowdsourcing platform Tencent SOHO,this paper combines clustering selection and worker pruning methods to effectively improve the success rate of tasks allocation and the quality of tasks completion in the platform,which in turn improves the benefits of the platform.Next,this paper studies the route planning problem in crowdsourcing platforms,considering the recommended task routes for workers from the workers’ perspective,ensuring the maximization of workers’ rewards,motivating workers to participate in more tasks,and at the same time improving the quality of task completion,which ultimately enhances the benefits of the platform.Finally,this paper studies the task filtering problem in crowdsourcing platform,and proposes using active learning algorithms to select the most valuable subtasks to be delivered to workers for the most common data labeling tasks in the crowdsourcing platform,so as to ensure that the requesters’ needs are met,minimize the task cost while ensuring the worker’s motivation,promote more participation of requesters and workers in the crowdsourcing platform,and then improve the platform’s benefits.For each of these problems,the corresponding algorithms are proposed,and detailed theoretical proofs and analyses are given,while full experimental validation is conducted on real and simulated data sets.