The residential real estate market in Lisbon: trend analysis and forecasting selling prices using machine learning and geographic information systems

Abstract

The real estate market has a significant effect on a country's economy. The study's objective was to analyze and predict trends in the residential real estate market in the city of Lisbon based on three target variables: Selling Price per Square Meter, Discount Ratio, and Absorption Period. The methodology used was CRISP-DM. First, a clustering method was applied to group the parishes into clusters with similar temporal characteristics. Then, three forecasting methods (curve fit, exponential smoothing, and forest-based) were used to predict future values for the target variables in each parish. The clustering results showed well-defined clusters for the Selling Price per Square and Discount Ratio, whereas the Absorption Period had poorly-defined clusters. The curve fit had the best performance in 61% of the cases between the three forecasting methods. One major advantage of our approach is that the target variables can be forecasted according to the best-fitted model for each parish. In addition, we propose a framework to provide a practical and versatile solution to forecast property sales. Also, the findings of this study provide useful information to stakeholders for policymaking, strategies for real estate market investments, and residential locations for sustainable urban development.