Title: From Zero to The Hero: A Collaborative Market Aware Recommendation System for Crowd Workers

Abstract: The success of software crowdsourcing depends on active and trustworthy pool of worker supply. The uncertainty of crowd workers' behaviors makes it challenging to predict workers' success and plan accordingly. In a competitive crowdsourcing marketplace, competition for success over shared tasks adds another layer of uncertainty in crowd workers' decision-making process. Preliminary analysis on software worker behaviors reveals an alarming task dropping rate of 82.9%. These factors lead to the need for an automated recommendation system for CSD workers to improve the visibility and predictability of their success in the competition. To that end, this paper proposes a collaborative recommendation system for crowd workers. The proposed recommendation system method uses five input metrics based on workers' collaboration history in the pool, workers' preferences in taking tasks in terms of monetary prize and duration, workers' specialty, and workers' proficiency. The proposed method then recommends the most suitable tasks for a worker to compete on based on workers' probability of success in the task. Experimental results on 260 active crowd workers demonstrate that just following the top three success probabilities of task recommendations, workers can achieve success up to 86%