Suitability Description And Dynamic Measurement Of Software Developer In Crowdsourcing

Crowdsourcing is an emerging business model that organizes distributed crowds to solve various problems by the Internet.Recently,this paradigm has also flourished to software development domain.With Internet,software crowdsourcing encourages the public to participate in the process of product development without limitations of time and space,to achieve a win-win situation of both company and workers.Existing crowdsourcing platform usually adopts a manual model for task assignment instead of providing automatic task allocation service.Based on the investigation of current related work,this paper proposes an approach to fast discover suitable workers for software crowdsourcing tasks.Our innovative research is important for the further promotion and implementation of crowdsourcing platform.First,the concept of “suitability” is put forward in this paper to describe whether a developer is suitable to accomplish a task.We believe that a comprehensive suitability description should be comprised of several dimensions: time,spatial,capability and price.With the suitability model,we can uniformly describe developer's information as well as task's constraints,thus the allocation process can be conducted quantitatively.Second,with the continuous growing number of developers and tasks,we adopt the clustering method to achieve efficient allocation process.Developers are divided into distinct clusters and the similarity between different clusters is small.Considering the special application background of crowdsourcing,we improve the traditional K-means algorithm and put forward a “twice division” clustering strategy.Our strategy effectively avoids the dimension disaster as well as the deviation of initial center selection.Third,we design a sliding window method to dynamically measure suitability description.This method adjusts the subjective deviations of self-evaluations by adopting objective evaluations given by task publishers.With the setting of parameters like the window size,the sliding distance and the historical weight,this method can maximize the accuracy and objectivity of measure results.Finally,in order to verify the correctness and effectiveness of proposed solution,we design and conduct some simulations to demonstrate the operability of suitability model,the efficiency and effectiveness of clustering strategy,and the accuracy of sliding window method.The results show that the allocation results are relatively reasonable and accurate,which greatly reduce the workload of both sides and ultimately realize the win-win goal.