Prediction of Real Estate Prices in Moscow using Hedonistic Models and Counterfactual Examples

  • Ђорђе Радивојевић Универзитет у Београду – Факултет организационих наука
  • Sandro Radovanovic Faculty of organizational sciences, University of Belgrade
  • Борис Делибашић Универзитет у Београду – Факултет организационих наука
Keywords: House price prediction, Hedonic price models, Counterfactual examples

Abstract

It is necessary to take into account several factors when predicting the price of real estate. More specifically,
in addition to the accuracy of the prediction, it is necessary to explain the prediction and to provide a mechanism that enables what-if analysis. However, the explanation of the decision and what-if analysis requires a cause-and-effect analysis, which is very complex by nature. In this paper, the procedure of prediction, explanation of the decision, and cause-and-effect analysis are provided by a combination of hedonistic models and counterfactual examples. Hedonistic models serve to predict and describe
how much the buyers value the properties of the real estate, and how much the properties of the real estate environment, while counterfactual examples give potential explanations why the buyer paid or did not pay for hedonism. Experiments are conducted on the purchase and sale transactions in the Moscow region, which describe each transaction with 20 properties, of which nine are properties of the property and 11 are properties of the real estate environment. Experiments show that it is possible to predict real estate prices quite successfully by applying hedonistic models with the root mean logarithm square error of the real estate
price is equal to 0.2364. Then, by applying counterfactual examples, it is established that there are three manifestations of hedonism on the market. These are access to transport, health and educational institutions, and proximity to additional educational and cultural institutions, and finally the hedonism of environmental quality.

Author Biographies

Ђорђе Радивојевић, Универзитет у Београду – Факултет организационих наука

Позиција: студент

Фирма: Универзитет у Београду – Факултет организационих наука

Јове Илића 154, Београд

Области:

Машинско учење, Базе података, Веб апликације, Клауд технологије

Борис Делибашић, Универзитет у Београду – Факултет организационих наука

Позиција: редовни професор

Фирма: Универзитет у Београду – Факултет организационих наука

Јове Илића 154, Београд

Области:

Системи за подршку одлучивању, Вишекритеријумско одлучивање, Пословна интелигенција, Машинско учење

Published
2022-02-08
Section
Information engineering