基于改进贝叶斯算法的云服务可靠性预测方法研究  

CLOUD SERVICE RELIABILITY PREDICTION METHOD BASED ON IMPROVED BAYES

作　　者：钱双洋 陈喆[1] 刘原序 

机构地区：[1]解放军信息工程大学,河南郑州450001

出　　处：《计算机应用与软件》2017年第11期34-38,69,共6页Computer Applications and Software

基　　金：国家高技术发展研究计划项目(2008AA01Z404);国防预研基金项目(9140A26010306JB5201)

摘　　要：目前服务消费者业务的不断发展,业务逻辑的不断复杂,对云服务组合的可靠性性能需求不断增加。通过对服务组合可靠性进行预测,根据预测结果向服务消费者推荐满足其可靠性需求的组合服务,以提高服务使用质量。基于传统的贝叶斯预测模型提出一种改进贝叶斯预测算法(IDLM)。通过采用指数加权回归方法对算法中的状态误差方差项进行估算,有效解决了传统贝叶斯模型中状态误差方差参数确定困难等问题,并且具有较高的预测效率和预测准确性。实验结果表明,改进的贝叶斯预测算法(IDLM)较其他传统的时间序列预测算法具有较高的准确性。In view of the continuous development of the service consumer business and the complexity of the business logic, the demand for reliability of the cloud service composition is increasing. Through the prediction of the reliability of service composition, according to the prediction results, the service composition was recommended to service consumers to satisfy their reliability requirements to improve the service quality. We propose an improved Bayesian prediction algorithm based on the traditional Bayesian prediction model. By using the exponential weighted regression method to estimate the variance term of the state error in the algorithm, the problem of difficulty in determining the variance parameters of state error in the traditional Bayesian model is solved effectively. And it has high predictive efficiency and predictive accuracy. The experimental results show that the improved Bayesian prediction algorithm has higher accuracy than other traditional time series prediction algorithms.

关 键 词：服务组合 可靠性 贝叶斯模型 预测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]