Abstract: Utilizing existing literature data on the split tensile strength of fiber reinforced coral aggregate concrete (FRCAC-SS),a database for FRCAC-SS was established.Four machine learning models for FRCAC-SS were proposed.The models’ performance was evaluated through a comparison of experimental and model values,model error distribution histograms,and five key performance indicators.The results indicate that the genetic algorithm-optimized support vector regression (GA-SVR) model closely approximates experimental values in both training and testing sets.The GA-SVR model exhibits smaller mean and standard deviation of error distribution,and it excels in all five performance indicators.The Shapely value reveals that FRCAC-SS is the most sensitive and positively correlated to cement.A graphical user interface was developed based on the GA-SVR model,enabling a visual representation of FRCAC-SS’s design.These findings have the potential to establish a foundational understanding for the application of FRCAC in reef-building projects.