Task Recommendation In Software Crowdsourcing Based On Developer's Dynamic Preferences And Competitiveness

In recent years,as a new schema of software development,software crowdsourcing has been widely concerned by academia and industry.Compared with the traditional software de-velopment,software crowdsourcing can make the most use of the developers all over the world to complete the complex development tasks,which can effectively reduce the development cost and improve the efficiency.However,due to there are a large number of complex tasks in the current crowdsourcing platform and inaccurate task matching will affect the progress and qual-ity of task completion,it is very important to study the matching problem between developers and tasks for software crowdsourcing.As an effective method to solve information overload in traditional fields,this paper aims to introduce recommendation technology into the matching problem between developers and tasks,that is,recommending the appropriate software development tasks for the developers in software crowdsourcing.The following aspects have be considered in this paper to solve this problem:on the one hand,developers choose tasks based on their preferences first,however,the developer's preferences are constantly changing and it is important to accurately capture the current preferences of developers;on the other hand,compared to the goods or other content in traditional recommendation fields,the software development task has the professional nature which requires people with corresponding skills to complete and meanwhile there are more com-petitive crowdsourcing platforms,so when choosing a task,developers also consider whether they have the ability to get a higher score among a wide range of competitors which this paper takes into account.In response to the above considerations,this paper studies and completes the following work:(1)the dynamic preferences and competitiveness are taken into account in modeling developers,and the parameters reflecting these two features are defined;(2)proposes a two-stage task recommendation model in software crowdsourcing:in the first stage,uses Attention Mechanism-based Long Short-Term Memory Network to predict the current dynamic preference of a developer,and employs similarity to screen out the Top-N tasks that conforms to the preference from a large number of candidate tasks;in the second stage,according to the developer's competitiveness,uses the improved Differential Evolu-tion Algorithm-based eXtreme Gradient Boosting to predict the developer's scores of tasks screened out from the first stage,and recommends the Top-K tasks to the developer accord-ing to the score from high to low;(3)to verify the validity of the proposed recommendation model,a series of experiments have been carried out to compare with the classical methods for recommendation and the recom-mended results under different parameters are considered.The experimental results illus-trate that the proposed model has significant advantages in task recommendation in software crowdsourcing.