The Optimum Moisture Content (OMC) is the moisture level at which soil acquires peak density during compaction, making it crucial in geotechnical engineering and construction to determine the optimal conditions for maximum soil strength and stability in infrastructure. The relationship between OMC and Maximum Dry Density (MDD) is integral, as both parameters together determine the compaction characteristics of soil. This correlation varies depending on various features, highlighting the combined significance of OMC and MDD in geotechnical applications. Current methods for determining OMC are costly and time-consuming, highlighting the need for more efficient approaches. Machine learning-based methods offer a promising alternative, facilitating the development of innovative predictive models and algorithms that can enhance the accuracy and efficiency of OMC predictions compared to traditional empirical techniques. This study employs the Multi-Layer Perceptron model to address challenges in constructing a machine-learning model. Additionally, it integrates two distinct meta-heuristic optimization approaches, namely the Coot Optimization Algorithm and the Golden Jackal Optimizer, to achieve optimal results. As a result of this integration, two hybrid models, MLGJ and MLCO, were generated in three layers of the MLP models. The productivity of the frameworks was assessed by drawing on five metrics, including R2, RMSE, MSE, SI, and RAE. The outcomes revealed that the MLCO (2) approach stands out as the top-performing predictor among the three models of this study. It acquires an impressive maximum R2 value of 0.998 in the training stage, indicating exceptional explanatory capability and exhibiting notably low RMSE and MSE values of 0.2 and 0.04, respectively. This approach signifies minimal prediction discrepancies as opposed to different models utilized in this study. On the other hand, MLGJ (1) was identified as the weakest model in this study, registering the lowest value for R2 at 0.965 and the highest value of RMSE at 1.568.
Chong Gao (Mon,) studied this question.