Scholarlink

<p>Within the MAPE-K (Monitor, Analyzer, Planner, and Executor-Knowledge) loop, the Analyzer plays a central role by assessing the adaptation requirement and refining the adaptation space for dynamic Self-Adaptive Systems (SASs). However, when faced with imbalanced or poorly featured data, Machine Learning (ML)-based analyzers often produce erroneous predictions, forcing human intervention. In this paper, we propose a hybrid approach that integrates a Generative Adversarial Network (GAN) with Automated Machine Learning (AutoML) to support reliable, accurate, and automated analyzers. Specifically, GANs are employed to enhance the training dataset by generating high-quality synthetic data. Complementarily, AutoML–through AutoGluon–is employed to automate model construction and hyperparameter optimization. In fact, the GAN addresses several data-related issues that often plague AutoML in dynamic settings. We validate the proposed approach on two variants of the binary DeltaIoT dataset. Experimental results demonstrate that our approach consistently achieves superior performance across relevant evaluation metrics and outperforms AutoML alone. To the best of our knowledge, this work is the first attempt to integrate GANs into an AutoML for MAPE-K Analyzer, providing an efficient and automated data-centric solution for SASs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An automated MAPE-K analyzer based on a GAN-AutoML approach

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Sabah Lecheheb,
  • Soufiane Boulehouache,
  • Said Brahimi

摘要

Within the MAPE-K (Monitor, Analyzer, Planner, and Executor-Knowledge) loop, the Analyzer plays a central role by assessing the adaptation requirement and refining the adaptation space for dynamic Self-Adaptive Systems (SASs). However, when faced with imbalanced or poorly featured data, Machine Learning (ML)-based analyzers often produce erroneous predictions, forcing human intervention. In this paper, we propose a hybrid approach that integrates a Generative Adversarial Network (GAN) with Automated Machine Learning (AutoML) to support reliable, accurate, and automated analyzers. Specifically, GANs are employed to enhance the training dataset by generating high-quality synthetic data. Complementarily, AutoML–through AutoGluon–is employed to automate model construction and hyperparameter optimization. In fact, the GAN addresses several data-related issues that often plague AutoML in dynamic settings. We validate the proposed approach on two variants of the binary DeltaIoT dataset. Experimental results demonstrate that our approach consistently achieves superior performance across relevant evaluation metrics and outperforms AutoML alone. To the best of our knowledge, this work is the first attempt to integrate GANs into an AutoML for MAPE-K Analyzer, providing an efficient and automated data-centric solution for SASs.