Research On Demand Forecasting And Ordering Strategy Of Cloud Resource Server

Cloud computing is an emerging computing method.With the continuous progress of computer technology and the continuous expansion of the Internet,it has gradually provided a more efficient way for people to obtain all kinds of information and handle all kinds of business.Among them,cloud resources are the virtual carriers of cloud computing.And cloud resources server as the physical carrier of cloud resources,at the same time with the Internet and manufacturing nature,its cost is expensive.Therefore,the core competitiveness of such enterprises mainly focuses on the accuracy of cloud resource demand prediction and reasonable ordering strategy of cloud resource server.With the continuous improvement of technology,cloud computing is applied more and more widely,and the demand data of cloud resources is huge,covering multiple features of various regions and products.It is difficult for the commonly used demand forecasting methods to find rules from it,so as to accurately predict future data.Machine learning can analyze and process large amounts of data.For the massive data of cloud resources,it can find the law of their demand,so as to predict future data.Through accurate prediction,the ordering strategy of cloud resource server can be optimized according to different service level demands,so as to effectively reduce operating costs and enhance the core competitiveness of enterprises.In this paper,using the usage data of cloud resources,the feature set is established,and four machine learning methods,namely random forest,support vector regression,gradient lifting tree,and integrated learning combined machine learning model,are selected to forecast the demand of cloud resources.It is verified that the machine learning prediction method is better than the classical prediction method,and the integrated learning prediction accuracy is better than the single model prediction accuracy.According to the cloud resource demand predicted by integrated learning,the ordering strategy of cloud resource server is optimized through the static ordering model and the static and dynamic combining ordering model and the change of service quality.Through experimental verification,static and dynamic combined ordering model is more conducive to the ordering of cloud resource server.The research in this paper shows that compared with classical prediction methods,machine learning methods are more suitable for discovering regularities from massive data.Similarly,the research results of this paper provide ideas for the subsequent demand forecasting and ordering decision optimization of products with the same dual nature of Internet and manufacturing.