Analysis of construction costs with artificial neural networks

In this thesis, for forecasting costs of multiple reinforced concrete residential buildings with Artificial Neural Networks (ANN), cost of construction of this kind of buildings has been calculated and used as data for an ANN. This network has a multi layer and back propagation structure with adviser to learn. Building elevations, unit numbers in a flat, normal flat areas, heights of flats, total flats, outer surface?s empty areas, outer surfaces total areas and average areas of the units in normal flats were assumed as mean criterias of the cost of each apartment. Result cost values calculated with ANN, has checked with the Unit Price Method (UPM) and Regression Analysis Method (RAM) to evaluate the performance of ANN. Using the data calculated with the ANN, building design parameters for minimum costs, has been determined. According to these results, it?s comprehensible that the results of ANN are nearer than the results of Regression Analysis to the real costs of these buildings. Using hybrid methods for solving this kind of problems, will be useful than using only one method. Studying with similar methods for calculating different kind of buildings costs, will create positive developments.