Optimal Quality Assurance Task Assignment for Freelance Crowdsourcing Applications

Crowdsource freelancing applications have been widely utilized by online marketplace users. They provide solutions to both freelancers and employers by working as a mediator between them. It is crucial that each freelancing task is assigned to the best suitable available worker since freelancing tasks are usually expensive and require domain knowledge. In this paper, we formalize the problem of the freelance task assignment as a multi objective optimization problem and proof that it is NP-hard. We propose a memetic based quality assurance task assignment algorithm that finds the most suitable top k-tasks for each worker in which the total quality of the crowdsourced output is maximized. The proposed memetic algorithm applies both differential evolution (DE) and Large Neighborhood Search (LNS) in a way to better make use of the exploration abilities of the differential evolution as a global search and the exploitation of the Large Neighborhood Search as a local search. The proposed memetic algorithm applies the differential evolution to find the best solution in each generation then the large neighborhood search (LNS) is used to search the neighborhood of that solution. Firstly, a list scheduling (LS) algorithm is used to convert the continuous vectors resulting from the differential evolution to the discrete vector that represent the initial solution to the large neighborhood search (LNS). Secondly, the exploration of the initial solution is done by the local search (LNS). The experiments are done using published standard test data from literature. The experimental results shows that the using a simple DE algorithm with the large neighborhood search as a local search yields the best performance.