Affiliations:
1Department of Economics, Mandakh University, Ulaanbaatar, Mongolia
2Management School, Mongolian University of Science and Technology, Ulaanbaatar, Mongolia
Accurately valuing residential properties is difficult in transitional economies because market data are fragmented, many transactions occur informally, and price trends are unstable. These conditions reduce the reliability of traditional appraisal methods. This study systematically compares nine machine learning models, covering regression, kernel-based, ensemble, and deep learning approaches, using a dataset of 9,326 housing listings from Ulaanbaatar, Mongolia. The methodology includes extensive hyperparameter tuning, five-fold cross-validation, and district-level validation to ensure robust findings. Model performance was assessed on a separate test set using R², Mean Squared Error (MSE), and Mean Absolute Error (MAE). The deep neural network (DNN) achieved the highest accuracy (R² = 0.918; MSE = 0.051), outperforming both XGBoost and Random Forest, while Support Vector Regression (SVR) showed the weakest results. The most influential price factors were district, total area, floor level, garage availability, and number of windows. Some interior characteristics, such as parquet or tile flooring, were linked to lower prices, suggesting a buyer preference for more modern designs. The study also presents a Docker-based web application for real-time price prediction, demonstrating the practical value of these models in settings with limited data. By examining Mongolia’s secondary housing market, this research offers new evidence on the potential of machine learning to improve transparency and support decision-making in real estate valuation.
Housing valuation, Machine learning, Deep neural network, Price prediction, Transitional economies
https://doi.org/10.21833/ijaas.2025.12.021
Sodnomdavaa, T., Gurbazar, E., & Sodnomdavaa, T. (2025). Application of machine learning algorithms in real estate valuation. International Journal of Advanced and Applied Sciences, 12(12), 237–243. https://doi.org/10.21833/ijaas.2025.12.021