Performance evaluation of SVM and GBM in predicting compressive and splitting tensile strength of concrete prepared with ceramic waste and nylon fiber

Waste management has become a new challenge for the construction industries since rapid urbanization is taking place worldwide. Ceramic waste is one such material which is being originated from construction sites and industries, imposing a significant risk to the environment due to its non-biodegradable nature. With the goal of waste utilization, this study aims to predict the compressive and splitting tensile strength of concrete made with waste Coarse Ceramic aggregate (CCA) and Nylon Fiber (NF) by using two distinct machine learning algorithms, namely, Support Vector Machine (SVM) and Gradient Boosting Machine (GBM). A comprehensive data set for testing and training the models containing 162 records of compressive and splitting tensile strength test results were considered from nine mix proportions. For training the dataset, parameters like cement content, sand content, stone content, ceramic content, nylon fiber content, curing duration, and concrete strength were taken as input variables. The predicted strengths obtained from the SVM and GBM based models are found to be in close agreement with the experimental results. In terms of coefficient of determination (R²), GBM showed significantly better result for both compressive strength (e.g., SVM Overall R² = 0.879 & GBM Overall R² = 0.981) and tensile strength (e.g., SVM Overall R² = 0.706 & GBM Overall R² = 0.923) prediction. Furthermore, based on the statistical accuracy measures like the mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), it has been observed that GBM has yielded much better performance compared to SVM in predicting the mechanical strength of concrete.