Publications

Abstract : In crowdsourcing systems, tasks are accomplished by a crowd of workers in a competitive mode. Since tasks are diverse in nature, workers face difficulties in selecting a task. This could be resolved by deploying a task recommendation mechanism. Existing methods for task recommendation do not exploit the participation of workers and their winning chances. Hence, the success rate of workers is low. This paper proposes a novel task recommendation scheme that utilizes the winning and participation probabilities along with the dynamic nature of crowdsourcing platforms. Further, we address the cold-start problem using a hierarchical mapping of skills. The proposed scheme is validated through simulation using real and synthetic data, in comparison with state-of-the-art recommendation methods. The results indicate that this model obtains competitive accuracy in task recommendation and improves the success rate of workers. We observed that the cold start problem reduces substantially with less computational overhead.