基于ARIMA-LSTM组合模型的应用系统响应时间预测

Response Time Forecasting of Application System Based on ARIMA-LSTM Hybrid Model

响应时间既直观反映应用系统的处理能力,又直接影响用户的服务体验,有效地预测应用系统的响应时间对提前感知并解决系统性能问题、提升用户满意度都有重要指导意义。有研究表明,组合预测模型比单一模型具有更好的拟合性,论文提出一种基于ARIMA-LSTM组合模型的应用系统响应时间预测方法。该方法根据响应时间的历史数据,先以ARIMA模型过滤数据中的线性趋势,再将残差传给LSTM模型以提取数据中的非线性特征,最后集成得到最终预测结果。系列实验结果表明,所提方法相比于单一预测模型具有更高的预测精度,拟合性能更优。

Response time in an application system is one of the most important performance metrics.It not only reflects the pro⁃cessing power of the application system,but also affects the user's intuitive experience directly.Effective prediction of response time has important guiding significance for early sensing and solving system performance problems and improving user satisfaction.Some studies have shown that the hybrid forecasting model has better fitting performance than the single model.Therefore,this paper pro⁃poses an application system response time forecasting method based on ARIMA-LSTM hybrid model.According to the historical da⁃ta of response time,the method firstly filters the linear trend in the data with the ARIMA model,and then passes the residual to the LSTM model to extract the nonlinear features in the data.Finally,the results of the first two steps are superimposed to get the final forecasting result.The experimental results show that the proposed method has higher forecasting accuracy and better fitting perfor⁃mance than the single prediction model.