Modeling Civil Engineering Problems via Hybrid Versions of Machine Learning and Metaheuristic Optimization Algorithms

The aim of this study is to develop hybrid solution models by integrating metaheuristic optimization algorithms and machine learning technique. These hybrid models are utilized to estimate bearing capacity of pile groups and lake levels, which are common challenges to calculate in the geotechnical and hydrology designs of civil engineering. To achieve this, Lake Beyşehir lake-water level observations and various pile group designs are used as a dataset, which is divided into two parts; 75% for training and 25% for testing. By employing improved hybrid models that combine metaheuristic algorithms such as harmony search, artificial bee colony, and particle swarm optimization with a machine learning technique called least squares support vector regression (LSSVR), optimal values of kernel parameters are obtained reliably and robustly. The results suggest that these hybrid models can be successfully applied to complex real-world problems, as evidenced by nine evaluation metrics, including mean absolute error (MAE), root mean squared error (RMSE), and determination coefficient (R2), which showed satisfactory and reasonable performance. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.