Abstract: The disjointing of cloud task-to-resource matching link is a prominent problem in the cloud fusion process. Aiming at this problem, a cloud-based fusion task assignment model optimization method based on the improved knowledge migration maximum entropy clustering algorithm (KT-MECA) is proposed, in which the two-sided satisfaction of tasks and resources is considered. The algorithm improves the introduction of historical clustering center knowledge and membership degree knowledge, improves clustering performance and stability, and solves the problem that traditional clustering algorithms cannot be applied to dynamic cloud resource clustering. Considering the two-side subject satisfaction, the result of KT-MECA is applied to the two-side matching decision optimization model of cloud task and resources. The example proves that this method is feasible.

Key words: cloud fusion, knowledge transfer, maximum entropy clustering, two-sided matching