基于社会关系的工作流任务分派策略研究

发布时间：2018-07-22 10:07

【摘要】：在工作流管理系统中,任务分派策略对工作流系统的性能影响较大,而人力资源社会属性的不稳定也给任务分派带来了挑战.一般的任务分派策略还存在以下问题:分派时只考虑候选资源的个体属性,忽略了流程中其他资源对候选资源的影响;需要为候选资源预先设置能力指标,但预设指标很难与候选资源的实际情况吻合,错误的能力指标会导致将任务分派给不合适的资源,降低工作流系统的性能.为克服上述问题,基于不同的状态转移视角和奖励函数,提出了4种基于Q学习的任务分派算法.通过对比实验,论证了基于Q学习的任务分派算法在未预设资源能力的情况下仍能取得较好效果,且支持在任务分派过程中考虑社会关系的影响,使得平均案例完成时间进一步降低.
[Abstract]:In workflow management system, task allocation strategy has a great impact on the performance of workflow system, and the instability of human resource social attributes also brings challenges to task allocation. The general task allocation strategy also has the following problems: only considering the individual attributes of candidate resources, neglecting the influence of other resources in the process on candidate resources, and setting capability index in advance for candidate resources. However, the default index is difficult to match with the actual situation of candidate resources, and the wrong capability index will lead to the assignment of tasks to inappropriate resources and reduce the performance of workflow system. In order to overcome the above problems, four Q-learning based task assignment algorithms are proposed based on different state transition perspectives and reward functions. Through comparative experiments, it is proved that the task allocation algorithm based on Q learning can still achieve good results without default resource ability, and supports the consideration of the influence of social relations in the process of task assignment. The average case completion time is further reduced.
【作者单位】：中山大学软件学院;中山大学信息科学与技术学院;
【基金】：国家自然科学基金(60873162) 广东省自然科学基金(S2012010009634) 广东省重大科技专项(2012A010800012) 广州市科技计划(12A12051586)
【分类号】：TP311.52

【参考文献】