Title: Predictive Modeling of Physical and Mechanical Properties of Pervious Concrete using XGBoost

Abstract: High permeability of pervious concrete (PC) makes it a special type of concrete utilised for certain applications. However, the complexity of the behaviour and properties of PC leads to costly, time consuming and energy demanding experimental works to accurately determine the mechanical and physical properties of PC. This study presents a predictive model to predict the mechanical and physical properties of PC using Extreme Gradient Boost (XGBoost). The compressive strength, tensile strength, density and porosity of PC was predicted using four models evaluated using different statistical parameters. These statistical measures are the root mean squared error (RMSE), square of correlation coefficient (R2), mean absolute error (MAE) and mean absolute percentage error (MAPE). The estimation of these properties by the XGBoost models were in agreement with the experimental measurements. The performance of XGBoost is further validated by comparing its estimations to those obtained from four corresponding support vector regression (SVR) models. The comparison showed that XGBoost generally outperformed SVR with lower RMSE of 0.58, 0.17, 0.98 and 34.97 compared to 0.74, 0.21, 1.28 and 44.06 in SVR for compressive strength, tensile strength, porosity, and density estimation respectively. Due to high correlation between the predicted and experimentally obtained properties, the XGBoost models are able to provide quick and reliable information on the properties of PC which are experimentally costly and time consuming. A feature importance and contribution analysis of the input/predictor variables showed that the cement proportion is the most important and contributory factor in the PC properties estimated.