The global crisis led to a significant decline in house prices. Financial institutions were the ones most affected, with major financial losses. The Greek market experienced an unprecedented situation regarding the current valuations and the future trends. The residential market in Greece has experienced significant contraction in the period 2010-2018.
In this project we used neural networks to value and forecast real estate prices in 240 different administrative sectors covering all areas in Greece. Special care was given for parameter tuning for neural network generalization improvement. More precisely, use the applied identification algorithms developed by Alexandridis and Zapranis (2013, 2014) and Zapranis and Refenes (1999). During the variable selection stage, we have selected only the statistically significant variables under a non-linear, non-parametric framework. Next, in the Model Selection stage, the network’s optimal architecture was selected again based on a statistical framework.
Our results indicate that the neural networks approach significantly outperforms alternative methods used by the financial institutions, and they highlight the importance of model selection in non-linear, non-parametric models.