基于隐马尔可夫模型的软件状态评估预测方法

Approach of Measuring and Predicting Software System State Based on Hidden Markov Model

随着软件系统功能和性能的强化和提高,企业的管理效率在不断提升,运营模式也越来越丰富.与此同时,软件系统变得越来越复杂,这向软件系统管理和维护提出了严峻的挑战.如何通过采集系统外部特征参数,对系统内部状态进行客观、准确地评估和预测,成为亟待解决的问题.为此,提出了一种基于隐马尔可夫模型的软件系统状态评估预测方法.该方法基于软件系统外在特征参数,通过K-means方法构建系统的观测状态,并以此建立隐马尔可夫模型,建立起系统外在状态(观测状态)和内部状态(隐藏状态)之间的联系;再利用三次指数平滑法对具有周期性变化的系统特征参数进行预测,即可预测系统未来状态.针对基于B/S软件架构的信息管理系统的实验,其结果表明该方法对系统状态评估和预测具有较高的准确性.

With increased improvement on the capability and performance of software systems, enterprises have improved the management efficiency and enhanced the business model. Meanwhile, as software systems become more and more complex, severe challenges for the management of software systems are encountered. This paper presents a method for measuring and predicting software system state based on hidden Markov model. It establishes the linkage between the exterior state（the observation state） and the interior state（the hidden state） of the software system. K-means method is applied to construct the observation state of system. Triple order exponential smoothing is used to predict the future state of the system exterior state which changes cyclically. The experimental analysis on B/S information management system shows that the proposed method has high accuracy for measuring and predicting the software system state.