In this study, a method is proposed for the preliminary estimation of school building costs. For this purpose, the projects of 96 school buildings, which were put out to tender by public institutions and the result announcement was published, were provided from the Electronic Public Procurement Platform. These projects are divided into 81 training data and 15 test data. The construction cost of educational buildings through projects may affect the output vectors; “Number of Classrooms, Construction Duration, Total Number of Floors, Floor Height, Building Height, Building Height Class, Basement Height, Earthquake Design Class, Floor Class, Ground Safety Stress, Bed Coefficient, Concrete Class, Number of Elevators, Wet Area, Raft Foundation Height, Floor Area, Basement Area, Total Area" parameters were determined and used in modeling and analysis. Analyzes were carried out with “SPSS Statistics 26” software. Using these parameters, Regression Analyzes (RA) were performed and equations were developed to estimate the costs of educational structures. Approximate Costs and Contract Prices were tried to be estimated with the developed models. The model in which all parameters were used with the created equations was the model with the best correlation level, with the determination coefficient R²=0.900 for the Approximate Cost Price and R²=0.927 for the Contract Price. An error rate of 17,5% was found between the approximate cost estimates obtained using the model and the actual costs. It was determined that there was an 18,2% error between the estimated contract prices and the actual contract prices. The Durbin-Watson criterion was used to check the consistency between predicted results and actual results. As a result, an approach that can estimate the approximate costs and contract prices of school buildings of different types and coefficients with error rates lower than 20% has been created. Both public institutions and contracting construction companies will be able to make realistic cost estimates by benefiting from these modeling, providing time savings. Conducting similar studies by increasing the number of data may be a solution to minimize the error rate in subsequent modeling.
Read full abstract