Scholarlink

<p>There is a significant need for optimization techniques in analyzing soil stabilized with waste-based stabilizer as an alternative to traditional methods. In this study, a hybrid metaheuristic algorithm (HMA) was employed to predict the maximum dry density (MDD) and unconfined compressive strength (UCS) using datasets derived from waste-based stabilized soils. Consequently, a golden eagle optimizer (GEO), hunger game search (HGS), weIghted meaN oF vectOrs (INFO) algorithm (INFO), rime optimization algorithm (RIME), red-tailed hawk algorithm (RTH), and Runge–Kutta algorithm (RUN) were employed for the task, hybridized with multilayer perceptron neural network (MLPNN) models. Metrics including mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and correlation coefficient (<i>R</i>) were employed to assess the accuracy of the models. The RUN model showed the highest accuracy in the training phase for MDD and UCS with RMSE of (0.0348, 0.0309), MAE (0.0224, 0.0232), MAPE (0.1034, 0.1360) and <i>R</i> (0.9976, 0.9985), while the INFO algorithm excelled in the testing phase of MDD and UCS prediction, with RMSE of (0.0875, 0.0848), MAE (0.0619, 0.0233), MAPE (0.3077, 0.1361) and <i>R</i> (0.9791, 0.9925). A predictive formula was developed to predict MDD and UCS, enhancing efficiency and reducing labor costs. However, due to data limitations, it is recommended that future research validate the findings with diverse datasets.</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

Prediction of Soil Parameters of Waste Fibers with Hybrid Metaheuristic Algorithm

  • 期刊论文

  • 发表日期：

    2026年2月1日

查看全文
  • Mohammed Abdullahi Mu’azu

摘要

There is a significant need for optimization techniques in analyzing soil stabilized with waste-based stabilizer as an alternative to traditional methods. In this study, a hybrid metaheuristic algorithm (HMA) was employed to predict the maximum dry density (MDD) and unconfined compressive strength (UCS) using datasets derived from waste-based stabilized soils. Consequently, a golden eagle optimizer (GEO), hunger game search (HGS), weIghted meaN oF vectOrs (INFO) algorithm (INFO), rime optimization algorithm (RIME), red-tailed hawk algorithm (RTH), and Runge–Kutta algorithm (RUN) were employed for the task, hybridized with multilayer perceptron neural network (MLPNN) models. Metrics including mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and correlation coefficient (R) were employed to assess the accuracy of the models. The RUN model showed the highest accuracy in the training phase for MDD and UCS with RMSE of (0.0348, 0.0309), MAE (0.0224, 0.0232), MAPE (0.1034, 0.1360) and R (0.9976, 0.9985), while the INFO algorithm excelled in the testing phase of MDD and UCS prediction, with RMSE of (0.0875, 0.0848), MAE (0.0619, 0.0233), MAPE (0.3077, 0.1361) and R (0.9791, 0.9925). A predictive formula was developed to predict MDD and UCS, enhancing efficiency and reducing labor costs. However, due to data limitations, it is recommended that future research validate the findings with diverse datasets.