Artificial Neural Network Models and Real Estate Market in EU Countries

The aim of our paper is to define a model for real estate price forecasts using artificial neural networks, based on a sample of 27 European countries. In order to create the model, we opted for 11 inputs: GDP, GDP per capita, inequality in income distribution, unemployment rate, labor productivity, FDI, HICP, VAT, property tax (as a % of GDP) and property tax (as % of total taxes). The training of the artificial neural network, using the advanced Backpropagation algorithm, was preceded by the collection and analysis of empirical data for 27 European countries. Of the 253 input data sets, 80% of the data were selected as the training set for the network, and 20% as the validation set. On 11 data sets, on which it was not trained, the network was controlled. Our research has shown that forecast models, based on the use of artificial neural networks, have a satisfactory degree of accuracy, so with a reliability of about 85% it is possible to apply a trained neural network for quick rough estimates of the price of apartments on the EU market.