Abstract:
Realistic estimation of construction cost is a vital issue for both successful planningand completion of every construction project. However, fluctuations in input prices due to the unexpected changes in factors like inflation and supply/demand balance make realistic costestimation very difficult to achieve. Thus, various estimation methods have been developedand these can be grouped as methods based on; statistics-probability analysis, comparison with similar projects and artificial intelligence techniques.Statistics-probability analysis is the most widely used method for construction costestimation in Turkey. Based on the so called method, Ministry of the Environment and Urbanism publishes and updates "Unit Costs of Construction" every year and the data is widely used for preliminary cost estimation by both the contractors and the developers.Meanwhile, methods based on artificial intelligence techniques are rarely used within the industry. Thus, the aim of this study has been to compare the estimation results obtained by using statistics-probability analysis and artificial intelligent techniques. In order to achieve this, construction cost data from 198 projects; completed between 2004-2010 in Izmir (the third largest city in Turkey) were used. Multi layer perceptron (MLP) and grid partitioning algorithm (GPA) were used to obtain estimation results and root mean square error (RMSE)and coefficient of determination (R2) were calculated for comparisons.
