Thesis (M.Sc.) - Cairo University - Faculty of Graduate Studies for Statistical Researches - Department of Bio-Statistics and Demography

The use of machine learning techniques is becoming more popular than classical statistical techniques due to its remarkable reduction of hypothesis. This advantage makes the machine learning techniques are increasingly used to simulate the behavior of concrete properties Which is an important research area.The compressive strength of the concrete is a major civil engineering issue.This thesis provides a comprehensive study using both classical statistical and machine learning techniques to predict the concrete compressive strength at different ages. Only multiple regression analysis is applied as a classical statistical technique while, different algorithms (linear regression, M5P tree and multilayer perceptron) are applied as machine learning technique.The variables used in the prediction models were from the knowledge of the mix itself. The independent variables are cement, fine aggregate,coarse aggregate, fly ash, BFS, free water and age while the dependent variable is the concrete compressive strength.To achieve this objective, an experimental program was divided into two phases: In Phase I, exploring process was planned to determine the most effective independent variables to the compressive strength in case of classical statistical model followed by obtaining the final model of the classic statistical technique. In Phase II, exploring, proposing models and obtaining the final model in case of applying machine learning algorithms followed by a comparison between all models of the two techniques.The results show that the Model developed using M5P tree Algorithm through machine learning technique is found to be the best model between the machine learning algorithms with an adjusted R Square value of 0.66 while, the model developed using Multiple regression analysis in case of applying classical statistical technique has an adjusted R Square value of 0.67 but it is used only to predict concrete without BFS and Fly ash.The M5P tree algorithms model verify that the machine learning algorithms model enjoys more flexibility,capability and accuracy in predicting the compressive strength of concrete

Issued also as CD