Scholarlink

The Composition and Adaptation of Ensemble Learning (CAEL), also referred to as the Harmonizing Ensemble Framework, is a highly effective approach for identifying Android malware. By combining multiple base learners, this method is able to adapt to different data distributions and complexities found in malware datasets. This collaborative model harmonization significantly improves the predictive performance of the framework. The process involves dataset splitting, feature subset selection, and training base classifiers such as MLP. The ensemble evaluation validates the strength of the models, with the ANN, LightGBM, and K-NN classifiers demonstrating well-balanced performance. Additionally, the SVM classifier achieves an accuracy of 85%. To further enhance performance, meta-learners dynamically adjust the weights of the base learners. The study highlights the reliability of various classifiers and utilizes the LIME algorithm for accurate feature interpretation, ensuring precise predictions. Notably, Instance 4 fulfills the necessary feature requirements, underscoring the significance of specific conditions for accurate predictions within the model.

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

Composition and Adaptation of Ensemble Learning for Android Malware Detection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • G. Kirubavathi,
  • R. N. Varun Vijay

摘要

The Composition and Adaptation of Ensemble Learning (CAEL), also referred to as the Harmonizing Ensemble Framework, is a highly effective approach for identifying Android malware. By combining multiple base learners, this method is able to adapt to different data distributions and complexities found in malware datasets. This collaborative model harmonization significantly improves the predictive performance of the framework. The process involves dataset splitting, feature subset selection, and training base classifiers such as MLP. The ensemble evaluation validates the strength of the models, with the ANN, LightGBM, and K-NN classifiers demonstrating well-balanced performance. Additionally, the SVM classifier achieves an accuracy of 85%. To further enhance performance, meta-learners dynamically adjust the weights of the base learners. The study highlights the reliability of various classifiers and utilizes the LIME algorithm for accurate feature interpretation, ensuring precise predictions. Notably, Instance 4 fulfills the necessary feature requirements, underscoring the significance of specific conditions for accurate predictions within the model.