Investigating the Effects of Mixing Design Variables on Performance of Asphalt Compaction Using Data Mining Algorithms

Many factors or parameters influence the field density of asphalt mixes. Since the number of variables affecting a large and somewhat different compacts affects each other, it is almost impossible to determine a constant interpolation relationship. Data mining and its techniques are a way of discovering hidden knowledge between dependent and independent variables, as well as indirect and nonlinear relationships can be identified by dividing the data into groups or leaves in the decision tree method. In this study, commonly used data mining techniques in civil engineering, including the neural network, logistic regression and decision tree. By emphasizing the application of decision tree method, with the aim of exploring knowledge models and providing predictions, other data mining tools are used to assist in constructing and evaluating the developed satistical model. The explanatory variables used in the three models of this study are the percentage of void, asphalt mix strength, aggregate size, bitumen percentage, asphalt mix flow. The results show that the percentage of void content of stone materials, the percentage of passage of sieve 200 and 4, and the bitumen percentage had a greater effect on the density and compaction of asphalt mixture. Also, a multiple linear regression model with a correlation coefficient of nearly one between the density of asphalt mixture and variables were presented.