A Computational Method to Predict the Concrete Compression Strength Using Decision Trees and Random Forest

Resumo

The engineering properties of concrete made structures depend on various parameters such as
the properties of the mixture of concrete, including its nonhomogeneous nature. A clear understanding of
such complex behavior is needed to use these materials successfully in various engineered structures.
Recently, the advancement of machine learning techniques has managed to propose different optimum
solutions to general engineering applications. This study aims to predict concrete compression strength by
employing methods such as Decision Trees (DTR) and Random Forests (RFR) using the database available
in Yeh [1]. The model used in this study considers the effect of eight contributory factors, i.e., cement, blast
furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate and age to predict the concrete
compressive strength. Computational methods like DTR and RFR are used to develop a predictive model.
A tuning method called GridsearchCV is also used to automate the process of adjusting the algorithms. The
study also compared the performance of the algorithms concerning their predicting abilities. The divergence
of the root means square error (RMSE) and R2 between the output and target values of the test set was
monitored and used to establish a better solution.