论文部分内容阅读
为保证基础设施即服务(IaaS)模式云环境中资源的有效分配与高效调度,提出了一种基于集成模型-优化神经网络的资源需求预测方法(EMONN).分析了用户偏好以及资源配置需求,根据需求变化采用阈值法确定波动期与平缓期,通过基本预测器集成模型实现不同时期、不同需求的预处理.预处理结果经过加权,与历史数据共同作为神经网络(NN)的训练数据,保证预测结果精度.为改善神经网络的学习率与稳定性,采用自适应学习率以及动量方法对神经网络进行优化.采用统计指标对系统有效性进行验证,结果表明所提方法可以精确有效实现用户需求预测.
In order to ensure the efficient allocation and efficient scheduling of resources in infrastructure as a service (IaaS) model cloud environment, this paper proposes a resource demand forecasting method (EMONN) based on integrated model-optimized neural network, analyzes user preference and resource allocation requirements, According to the change of demand, the threshold value method is used to determine the period of volatility and the flat period, and the basic predictor integration model is used to realize the pretreatment of different periods and different needs.The preprocessing results are weighted and used together with the historical data as neural network (NN) training data to ensure To improve the learning rate and stability of neural network, adaptive learning rate and momentum method are used to optimize the neural network.Using statistical indicators to verify the effectiveness of the system, the results show that the proposed method can accurately and effectively meet user needs prediction.