Research And Implementation Of Recommendation System Based On Microservices And Workflow Technology

The recommendation system for radio and television users involves a lot of business scope,such as film,television and news.Due to different business scope and user characteristics,some demands are generated.First of all,different recommendation strategies should be adopted according to users' needs under different contents.Therefore,it is necessary to divide and configure these services into multiple levels and modules.Secondly,if the above multiple business scenarios are realized,the business process logic of the system will be more complex.The common implementation method is hard coding,which makes it difficult for administrators to clearly view the dynamic business process.The third problem is caused by the large fluctuation in the computing resources of the radio and television program recommendation system,which leads to the low utilization efficiency of resources and the negetive impact on the recommendation effect.Based on the above background,this thesis focuses on the business scope division,business flow management and improvement of recommended computing amount under limited resources.Then the system was implemented by using Recommendation technology,micro-service,workflow and hot update container deployment technology.Firstly,this thesis constructs a management system for business.System administrators can manually add business modules and strategy algorithms to the management system.Then these business modules and strategy can be combined through flexible configuration to form a variety of recommendation scenarios.In terms of the implementation of recommendation scenarios,based on the three layers architecture of the recommendation system,including offline,online and near-line,the relevant recommendation,personalized recommendation and hot recommendation scenarios are realized.Secondly,this thesis uses the workflow engine to manage the business process.The workflow engine applies the business process modeling symbol standard to describe the workflow,which makes the business process more clear and easily controls the dynamic change of the business.The traditional workflow is realized by single application,in this thesis,micro-service is used as a single task of workflow engine to realize the change of system deployment.Finally,as a supplement,in order to increase the amount of recommendation computing under limited resources and thus improve the overall recommendation effect,this thesis adopts the dynamic hot update design of the deployment layer.That is,during the operation of the recommendation system,the deployment state of the near-line layer and the online layer is dynamically updated according to the change of system load.The flexible configuration of services makes the system more adaptable to different recommendation requirements and enables the system to realize various application scenarios based on the fixed recommendation computing architecture.Through the workflow,the business process can be visualized and standardized management,which makes the static composition and dynamic change of the recommended business process more clear and intuitive.The combination of workflow and micro-service facilitates the deployment and extension of the recommendation computing task,which makes the deployment of the system have the characteristics of dynamic change.Then the recommendation computing amount of different scenarios is always maintained at a high level,and the recommendation effect of the system is improved.