Methods of automated assessment of commercial real estate

Automated valuation models (AVMs) have been the subject of study for many decades. This paper discusses some of the main methods that have been used in the literature, and aims to reveal their advantages and limitations for solving the problem. First, some of the more traditional techniques are considered, namely the hedonic prices regression model (HPM) and geographically weighted regression (GWR). Further, “newer” machine learning algorithms are considered, namely decision trees, random forest (RF - random forest) and gradient boosting (GBM - gradient boosting model) Finally, these methods are evaluated based on the potential accuracy of prediction to evaluate individual properties and the degree of interpretability and reliability of the results.