Research On Time Series Data Collaborative Prediction Algorithm In Computing Power Constrained Environment

Time-series data forecasting algorithm has become the basis of a lot of social service,and use machine learning to forecast time-series data is a common means of when need a large amount of computing resources will be delivered to cloud computing tasks for processing platform,cloud platform for algorithm provides abundant force support,cloud platform,however,lack of flexibility and consumes too much bandwidth.Edges end,this algorithm can be sinking to the lightweight can overcome the shortcomings of cloud platform clumsy,but the edge of the end because of the limitation of its hardware is like a cloud platform provides abundant strength,cause in the time-series data to predict when there will be a shortage of work force,therefore,in this thesis,the introduction of collaborative computing,with the two kinds of typical time-series data forecasting algorithm combining,A collaborative prediction algorithm of time series data is formed to reduce the computing cost and memory consumption of the edge,and reduce the time-consuming of the algorithm.The specific work is as follows:(1)Summerize the preliminary knowledge,especially the main theories of collaborative computing,elaborate the collaboration among cloud,edge and object at the end level,and elaborate the collaboration among resources,tasks,models and data at the computing level,and establish the theoretical framework.(2)This thesis combs the execution process of the short-term memory prediction algorithm and the time canonical matrix decomposition algorithm,analyzes and counts the calculation amount and storage requirements of each step,and finds out the module that consumes the most computing resources and has the highest computing density.(3)Collaborative computing is introduced,and the edge end and the cloud are taken as the collaborative objects.During calculation,the collaborative strategy is formulated according to the real-time supply of resources at the edge end and the computing resources required by the current computing module.(4)The collaborative prediction algorithm is simulated and the consumption and time consumption of computing resources are counted at the edge end.The experimental results show that the long and short term memory prediction algorithm and the time regular matrix decomposition algorithm with the synergistic mechanism can help the edge end save a lot of memory consumption and reduce the time consumption.At the same time,since the implementation logic of the algorithm is not changed,the synergistic mechanism does not affect the final output result.The time series data collaborative prediction algorithm realizes the collaborative computing between the cloud and the edge.Through to alleviate the pressure of the edge of the calculation,the algorithm makes full use of the cloud of idle computing resources,alleviate the resources and algorithm of edge side can provide the needed resources during the execution of a mismatch between the contradictions,improve the efficiency of the use of the cloud,system reliability,reduce the use of bandwidth,enables the edge of the end to faster response to task.