Since software systems become more complex than before, software ageing problems have a big impact on the performance of running software systems. To find software ageing in advance, some prediction methods were used to forecast those parameters which can indicate software ageing occurrences. Since the unsuitable parameters can reduce the prediction ability of an algorithm, in this study, multilayer perceptron (MLP) with ridge is proposed to improve the prediction accuracy of MLP and apply in software ageing problems. The proposed approach is a three-step method. First, a pre-processing process needs to be done by using outlier recognition, dispose, and normalisation. Second, MLP with ridge is proposed and used to optimise network structure. Third, a glowworm swarm optimisation method is utilised to automatically find optimal values of model parameters. In the experimental section, the results indicate that the proposed algorithm owns higher forecast accuracy than other state-of-the-art methods on two levels.

中文翻译：

---

使用带有岭的神经网络的软件老化预测

由于软件系统比以前更加复杂，因此软件老化问题对正在运行的软件系统的性能有很大影响。为了提前发现软件老化，使用了一些预测方法来预测那些可以指示软件老化发生的参数。由于不合适的参数会降低算法的预测能力，因此，本文提出了一种带有脊的多层感知器（MLP），以提高MLP的预测精度，并将其应用于软件老化问题。所提出的方法是三步方法。首先，需要通过使用异常值识别，处理和规范化来完成预处理过程。其次，提出了带有岭的MLP并用于优化网络结构。第三，利用萤火虫群优化方法自动找到模型参数的最佳值。在实验部分，结果表明，与其他现有技术相比，该算法在两个层次上具有更高的预测精度。