Prediction of Compressive Strength of Concrete Using M5’ Model Tree Algorithm: A Parametric Study

Abstract

Concrete mix proportions proposed by empirical equations do not account for slight modifications. Mix design becomes primitive in case of introduction of new parameters as research progresses such as accounting for new kinds of admixtures, super-plasticizers or binders. In practice and theory, effect of age on compressive strength is correlated. Metric of concrete compressive strength with age is an important criterion in prediction problem. Prediction problem parameterisation as quantity or ratios is a controlling model choice decision. The existing codes of practice do not account for a standardised process of evaluating compressive strength. Considering the strength prediction problem here as a classification domain of input space, it is modelled using M5’ model tree algorithm. The study conducted shows the performance promised by such a model to be accurate within statistical error.