Please use this identifier to cite or link to this item:
|Title:||Stable task assignment for crowdsourcing||Authors:||Li, Zefeng||Keywords:||DRNTU::Engineering::Electrical and electronic engineering||Issue Date:||2018||Abstract:||Task assignment is an important issue of crowdsourcing. The objective of this issue is to produce a stable and feasible task assignment. Stable means both workers and tasks are satisfied with their allocation and feasible means the reward for workers doesn't exceed the budget of tasks. Some existing works achieve this goal by using deferred acceptance method. But all these works assume that each worker has the same reliability towards different tasks for simplicity. Hence, no matter which task the worker is assigned to, the reward he receives is the same since the reward is positive correlated to reliability. As a result, workers only need to choose tasks according to their preference. However, in real life, the most common circumstance is each worker has heterogeneous reliabilities to different tasks due to their diverse education and life experiences. Under this circumstance, the existing algorithms may not be suitable anymore. Therefore, a new method is required to produce stable and feasible task assignment under this circumstance. In this dissertation, a revised algorithm is proposed based on the previous work to produce stable and feasible task assignment when the reliabilities of each worker are heterogeneous to different tasks. The key idea is to find a method for workers to let them choose tasks according to both their preferences and rewards. This algorithm can make better use of the worker's reliability because the worker will take reward as a consideration when choosing a task. The stability and feasibility of this algorithm are also proved. This dissertation makes comparisons between the revised algorithm and two existing assignment algorithms. Simulation results show that the revised algorithm proposed in this dissertation achieves a higher total task quality than the previous algorithms at the expense of workers' happiness. At the same time, workers will get more paid by using this revised algorithm.||URI:||http://hdl.handle.net/10356/76377||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.