Artificial Neural Networks Application for the Croatian School Maintenance Cost Estimation

The quality of education is associated with the condition of the infrastructure in which the educational process occurs, necessitating the continuous maintenance of these facilities. Limited and often insufficient maintenance funds pose a challenge in this context. Current cost estimates are inaccurate, and data on school and maintenance costs are constrained. More accurate maintenance cost plans would contribute to a better understanding of budget distribution and more efficient financial management. This study aims to investigate the application of artificial neural networks (ANNs) in planning annual maintenance costs for schools in the Republic of Croatia (Primorje-Gorski Kotar County). Using a database and DTREG software, three different ANN models were developed: a multilayer perceptron (MLP), a generalized regression neural network (GRNN), and a radial basis function neural network (RBFNN). Comparisons of the results showed that the GRNN is optimal and achieves the highest accuracy in estimating school maintenance costs. These findings can benefit educational institutions and public bodies in budget planning and decision-making regarding maintenance.