Abstract:  Effective load forecasting is of great significance to improving the utilization efficiency of cloud computing resources, reducing the energy consumption of cloud computing centers and reducing operation, and maintenance costs of cloud service providers Aiming to the problem of low accuracy of cloud computing resource load forecasting and inaccurate grasp of the time series characteristics of historical data, a combined forecasting model of BPNN_LSTM with time series factors is proposed Considering the date and time factors, which have significant influence on cloud computing resource load, LSTM is used to correct the residual of BPNN Compared with other prediction models, the proposed model has higher accuracy and stability in cloud computing resource load forecasting

Key words:  cloud computing resources, load forecasting, time series factors, BP neural network, long-term and short-term memory network