Combined Forecasting Schemes Based on Random Forest Regression and Meta-heuristic Algorithms for Maximum Dry Density

Compression is the method of compressing soil by decreasing the amount of air space within it. The level of compression needed for a particular soil is determined by its dry density, which reaches its highest point when the soil has the ideal moisture content. Nonetheless, accurately determining maximum dry density (MDD) through laboratory testing is time-draining and laborious. Therefore, combined forecasting schemes based on meta-heuristic optimization can reduce the required effort. The development of blended schemes was done using Random Forest Regression (RFR) along with three different meta-heuristic schemes, namely the FDA, GJO algorithm, and Gorilla Troops Optimizer (GTO). In this exploration, 30% of the blended schemes developed were deployed for training, and 70% were testing schemes. Many quantitative appraisal factors, including R2, RMSE, MDAPE, MAE, and U95, were deployed to examine the accuracy and effectiveness of the configured schemes. All the created schemes, particularly RFGJ—a hybrid of the RFR and the GJO algorithm—exhibited exceptional performance in predicting MDD values, attaining the top R² of 0.9966 and the least RMSE of 13.688.