基于分组学习粒子群算法的众包软件项目调度

Crowdsourcing software project scheduling based on group learning particle swarm optimization algorithm

为解决众包软件项目调度问题中的开发者选择、任务分配和投入度确定3个强耦合子问题,引入开发者信誉度,考虑技能、工作时长、开发团队规模等约束,以项目完成质量和工期为目标建立数学模型。提出一种采用三段式混合编码的分组学习粒子群算法求解所建模型。所提算法根据适应度排序将种群划分为3组,不同分组的粒子数量随进化代数自适应变化,且各组根据不同的适应度采用不同的更新策略。将所提算法与10种具有代表性的算法在12个不同规模的众包软件项目调度算例中进行对比,结果表明,所提算法能够获得精度更高的调度方案。

To solve the three coupled subproblems of the crowdsourcing software project scheduling including developer selection,task assignment and determination of the dedications,by introducing the reputation of the developers and considering the constraints such as task skills,working hours and team size,a mathematical model was constructed aiming to maximize the completion quality and minimize the project duration simultaneously.A group learning particle swarm optimization algorithm was proposed to solve the model,which adopted a three-segment hybrid encoding method and divided the population into three groups according to the fitness ranking.The number of particles in different groups changed adaptively with the evolutionary generation,and each group employed distinct update strategies according to the differences of fitness values.The proposed algorithm was compared with 10 representative algorithms on 12 instances with different scales.Experimental results showed that the proposed algorithm could obtain a scheduling solution with higher precision.