Abstract

Reliable AI methods for predicting and formulating relationships between input and output variables are particularly valuable when making data-driven decisions. However, conventional AI methods are primarily “black-box” techniques that lack interpretability. To address this issue, a novel, supervised machine learning model known as Symbiotic Operation Forest (SOF) is introduced. SOF improves the existing model using three strategies, including: (1) enlarging the geometric structure of the operation tree (OT) for better generalization, (2) using a bootstrap sampling technique to construct a diverse OT as a base learner, and (3) implementing a chaos-based algorithm using symbiotic organisms search 2.0 (SOS 2.0) to minimize the prediction error. Experiments were conducted on three datasets covering three different concrete types, including, respectively, recycled aggregate concrete (RAC), high-performance concrete (HPC), and lightweight foamed concrete (LFC). SOF was compared against several commonly used AI models using metrics such as RMSE, MAE, MAPE, R, and R2,with the results showing SOF to have both the best predictive performance and the capability to generate explicit formulas. The statistical results also confirmed the predictive performance of SOF to be significantly better than these AI models, with a p-value < 0.05.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call