基于LSTM的WEB服务响应时间大数据预测方法

Big data prediction method of WEB service response time based on LSTM

摘要:

有效地预测Web服务器响应时间,对Web服务提供方保障服务质量有着重要的指导意义.利用大数据方法对处理大量历史指标数据的处理能提高预测的效率.本文提出一种使用相关性分析除去与响应时间相关性不高的指标项,使用特征降维的方法减小计算的数据量,使用动态调节参数的多层LSTM优化算法对数据做训练并预测响应时间的方法来提高预测的效率和准确率.通过实验证明,本文提出的方法能高效和准确地预测Web服务响应时间.

Abstract:

Effective prediction of web services response time has important guiding significance for service providers to guarantee the quality of service, using big data approach to process a large number of indictors’ historical data can improve the efficiency of prediction. Correlation analysis is proposed to remove indicators’ items that are not highly correlated with response time, the computed data volume is reduced by the feature dimension reduction, and the multilayer LSTM optimization algorithm with dynamic adjustment parameters is designed to predict the response time of web services. Experiments show that the proposed method can predict the response time of Web services efficiently and accurately.