Application of Red Fox Optimization-Based Estimation Algorithms to the Flexural Strength of Basalt Fiber- Reinforced Concrete

In the current research, two approaches known as Support vector regression (SVR) and Adaptive neuro-fuzzy inference system (ANFIS) were studied to do target estimation called flexural strength (FS) of basalt fiber reinforced concrete (BFRC). This experiment’s accuracy depends on hyperparameters optimized using Red Fox Optimization (RFO) with SVR and ANFIS to find the best results (abbreviated SVRRFO and ANFISRFO). A database of 245 samples, with ten input parameters related to concrete components and fiber properties, was analyzed. Additionally, a sensitivity analysis was conducted to evaluate the impact of removing each variable on the target parameter. This study evaluated two machine learning algorithms’ prediction potential for basalt fiber reinforced concrete’s FS and clarified their key concepts. Removing fine aggregate and fiber content variables from the input set significantly increased Theil inequality coefficient (TIC) and decreased coefficient of determination (R^2) value. In terms of R^2 values, ANFIS_RFO outscored SVR_RFO (0.9802 and 0.9705) with 0.9859 and 0.9907, respectively. With lower values of error-based metrics in the learning and assessment portions, the discrepancies in the outcomes of the two schemes demonstrate that ANFIS RFOR is better than SVRRFO. The bulk of the input parameters (cement, fly ash, silica fume, coarse aggregate, water, water reducer, fiber diameter, fiber length, and fine aggregate) have only a somewhat unfavorable impact on the outcome, where removing fiber content variable from the input set significantly increased TIC and decreased R^2 value. This highlights the reliability and power of ANFIS RFO over SVR RFO, despite SVR_RFO’s acceptable accuracy.