Abstract:
Utilizing silica fume as a substitute for cement in pervious concrete offers a viable approach to achieve sustainability within
the realm of construction industry. The mechanical characteristics of pervious concrete are influenced by several factors,
such as quantity of silica fume utilized as a replacement for cement, the cement content, coarse aggregate, sand, admixture
and water used in the mix, aggregate size and the curing period. The present study introduces a predictive model that utilizes
machine learning approaches to estimate the compressive strength of pervious concrete blended with silica fume. The
models underwent training and testing procedures using 222 datasets from various literature sources. In this study, seven
machine learning algorithms were used as statistical evaluation methods to identify the most suitable and reliable model for
predicting compressive strength of pervious concrete. Among several models under consideration, the eXtreme Gradient
Boosting model showed superior performance in forecasting compressive strength of pervious concrete. The coefficient of
determination value obtained for training data is almost one, which suggests a robust correlation between the anticipated
and actual values. The root mean squared error of training data is 0.28 MPa, which indicates the mean variation between the
predicted and observed values. The coefficient of determination value for the test datasets is 0.97, along with a root mean
squared error of 2.21 MPa. The outcomes of the sensitivity analysis conducted on the eXtreme Gradient Boosting model
indicate that the parameter with the most significant impact on predicting the compressive strength of pervious concrete is the
admixture content, followed by the curing period. This work provides a comprehensive evaluation of the compressive strength
of pervious concrete, thereby enhancing the existing knowledge and facilitating its practical application in this domain.
