Property Valuations in Times of Crisis: Artificial Neural Networks and Evolutionary Algorithms in Comparison

In the current economic situation, characterized by a high uncertainty in the appraisal of property values, the need of “slender” models able to operate even on limited data, to automatically capture the causal relations between explanatory variables and selling prices and to predict property values in the short term, is increasingly widespread. In addition to Artificial Neural Networks (ANN), that satisfy these prerogatives, recently, in some fields of Civil Engineering an hybrid data-driven technique has been implemented, called Evolutionary Polynomial Regression (EPR), that combines the effectiveness of Genetic Programming with the advantage of classical numerical regression. In the present paper, ANN methods and the EPR procedure are compared for the construction of estimation models of real estate market values. With reference to a sample of residential apartments recently sold in a district of the city of Bari (Italy), two estimation models of market value are implemented, one based on ANN and another using EPR, in order to test the respective performance. The analysis has highlighted the preferability of the EPR model in terms of statistical accuracy, empirical verification of results obtained and reduction of the complexity of the mathematical expression.