Expert Resource Patterns Matching For Topic Vector

As a kind of shared resource in cloud environment,review experts providing academic review services have been widely used in a variety of service platforms.How to efficiently allocate academic review tasks to appropriate peer experts to maximize task matching quality and load balancing among expert resources is an important problem to academic review managers,The main challenges include:(1)the content of review tasks and experts’ research interests are usually described in text.The matching quality of review tasks is measured by the text similarity between task content and experts’ research interests.How to accurately describe the text similarity model is the key to optimize the quality of task matching.(2)Due to the limited expert resources,the task with high level priority can be matched with experts with high similarity to their review topics.Due to the constraint of the maximum number of acceptable tasks of expert resources,the review tasks assigned to expert resources later often have low matching degree with their research interests.How to reasonably match the review tasks with expert resources to maximize the matching quality is a difficult problem.(3)Because task is expected to be assigned to the expert resources with high matching degree,it is easy to cause the overload of expert resource,resulting in the queuing of relevant tasks and exceeding their deadline.How to balance the matching quality of task and the load of expert resources is also a challenge.For the allocation of expert resources in cloud environment,this thesis analyzes the constraints of academic review tasks on expert resource demand,task deadline and the maximum number of acceptable tasks of expert resources,and proposes the system framework and mathematical model of the problems considered with the optimization goal of task matching quality and the load of expert resources;Based on the classical topic model,the topic vector of review tasks and expert resources is extracted;In order to improve the matching efficiency,a k-means topic vector clustering model is proposed to mine task patterns and service patterns with similar attributes;A heuristic algorithm framework for topic coverage incremental aware resource matching based on service pattern is proposed,including task pattern scheduling sequence generation,service selection,resource matching,allocation result adjustment and other algorithm components.Two scheduling sequence generation rules are proposed,which are maximum average matching degree first and maximum topic coverage demand first;Three service resource balance selection rules are proposed: minimum load first,maximum acceptable number of tasks first and maximum average matching degree first;Two resource matching rules are proposed:maximum topic matching degree first and maximum topic coverage increment first;After resource allocation,appropriate resources reassign to tasks with insufficient allocated resources.In order to verify the performance of the proposed algorithm,the multi-factor analysis of variance(ANOV A)is used to correct the relevant parameters and components of the algorithm and select the best combination of parameters and components.The algorithm proposed in this thesis is compared and analyzed with the classical algorithms of other two related problems.The experimental results show that the performance of the proposed algorithm is better than the comparison algorithm under the conditions of different types of tasks and different demand to the number of resource.