Explainable models to estimate the effective compressive strength of slab–column joints using genetic programming

In this study, explainable models to predict the effective compressive strength of slab–column joints were proposed using genetic programming (GP). Effective compressive strength test results were collected from the literature, among which 81 were sandwich, corner, and edge columns, 44 were interior columns. With respect to the collected test results, random forest (RF), random forest–recursive feature elimination (RF-RFE), K-means clustering and linear regression analysis (KL) were applied to analyze the key features influencing the effective compressive strength. In addition, explainable models to predict the effective compressive strength were derived by using GP. The proposed models were verified by comparing with the existing equations, RF, and artificial neural network (ANN). Using the generative adversarial network (GAN), 20,000 fake data were additionally generated to strengthen the verification stage. The results showed that the proposed models exhibited the lowest root mean squared error (RMSE) among the existing equations, and RMSE of the proposed models was close to that of RF and ANN despite its conciseness.