Selected public Greek highway projects are examined in order to produce models to predict their actual construction cost based on data available at the bidding stage. Twenty highway projects, constructed in Greece, with similar type of available data were examined. Considering each project’s attributes and the actual cost, correlation analysis is implemented, with the aid of SPSS. Correlation analysis identified the most predictive project variables. Additionally, the WEKA application, through its attribute selection function, highlighted the most efficient subset of variables. These selected variables through correlation analysis and WEKA are used as input neurons for neural network models. FANN Tool is used to construct neural network models. The optimum neural network model produced a mean squared error with a value of 7.68544E-05 and was based on budgeted cost, lowest awarding bid, technical work cost and electromechanical work cost.