Prediction of residential real estate selling prices using neural networks

Abstract: With the rising housing prices of the last 20 years, the appraisal of real estate has become more difficult. Underlined by the large differences between listing and selling prices, the valuation process brings a level of uncertainty. With the advances within the field of machine learning in recent years, attempts have been made to apply these techniques to the real estate market. This thesis investigates the potential of using neural networks to predict selling prices of apartments in Stockholm, based on apartment parameters. Networks are trained to either make an improved valuation, based on a listing price, or make a new valuation of an apartment. The results are promising, and in line with contemporary findings; however, the worst-case performance of the models could make them unsuitable for many purposes.