Solving Team Making Problem for Crowdsourcing with Evolutionary Strategy

Crowdsourcing has become a popular service for requesters to integrate the human intelligence to complete software task. Since Crowdsourced Virtual Teams (CVT) are the foundation of the success of crowdsourcing tasks, previous studies have proposed various algorithms to solving CVT problem, including Alternating Variable Method (AVM), Hybrid Meta-heuristic algorithm (ES-AVM), etc. However, there is still room for improvement in performance. In this study, we propose to apply Evolutionary Strategy algorithm with Self-Adaptation (ESSA) to help publishers identify ideal CVTs. ESSA is effective which leverages self-adaptation mechanism to search solutions. We experimentally evaluate ESSA with 6,000 random classic instances. ESSA achieves the state-of-the-art results. Due to the lack of open databases to construct instances, we additionally construct a dataset with 1,556 realistic instances for the CVT problem. Experimental results show that ESSA significantly outperforms AVM and ES-AVM over 1,527 and 717 of the 1,556 realistic instances respectively.