A comparative analysis of decision tree on torsional capacity of reinforced concrete beams

In order to design or evaluate framed concrete structures subjected to eccentric loadings, it might be necessary to estimate one property of reinforced concrete (RC) of beams called torsional capacity (Tr) with sufficient accuracy. Unfortunately, current prediction algorithms are still not able to forecast the Tr of RC beams exactly, especially for over-reinforced and high-strength RC beams. By developing accurate estimation models, this drawback may be addressed as an alternative to other more complex and computationally demanding models. This study aims to evaluate and determine the best tree-based machine learning (ML) techniques for estimating the Tr of reinforced concrete beams under pure torsion. A computer program called Decision tree (DT) analysis was used as a tree-based technique to achieve this goal. DTs are widely known for their simplicity, interpretability, and efficiency in handling nonlinear relationships between variables. However, the performance of DTs is highly dependent on how well the decision thresholds (or splits) are optimized. In this study, the combination of DT with the Gannet optimization algorithm (GaOA) and Bald Eagle Search (BaES) introduces a novel way of addressing this challenge. Two hundred two participants a suitable percentage, were chosen from the literature and included in the training and testing phases of the data collection. Based on the results, the DT-Ba approach demonstrated great practical dependability with R2 values of 0.9896 and 0.9918 during learning and assessment. DT-Ga outperformed DT-Ba in R2 values, with values of 0.9964 and 0.9922, following the pattern.