Advancement of automated valuation model for commercial real estate: the impact of urban development investment sentiment

Abstract

This paper presents an ’automated valuation model (AVM)’ to automate the time and cost issues inherent in the traditional valuation methodology by the experts. In the process of implementing regression models for estimating real estate prices, existing studies have dealt with three main categories of independent variables: architectural, urban and macroeconomic features. On the other hand, real estate is an asset that is sensitive to local urban development, and the announcement of urban development alone effects the investment sentiment. However, there are not many AVM studies that have examined urban development investment sentiment (UDIS) as an independent variable. Therefore, this study examines how UDIS affects the price estimation of AVM. As a research method, the investment sentiment according to the urban development stage was defined to fit the regression model, collected and preprocessed the data reflecting it, and trained the data into five regression models including multiple linear regression and machine learning to examine the effect of treatment of UDIS variable. Furthermore, the contribution of the UDIS variable was evaluated to understand the internal dynamics regarding the feature. The results showed that the models with UDIS exhibited on average 0.73% more accuracy accross all dataset-model combinations. And SHAP analysis, an ad-hoc explainable model, confirmed that machined learning models also reflect the phenomenon that the closer the transaction is to the announcement of urban development, the more it affects the price. The significance of this study is that it validated the validity of market sentiment as an independent variable, which has been rarely addressed in AVM studies. Furthermore, this study unveils the development method of private enterprise lead commercial AVMs via academic means, which opens up possibility for it to develop further and reduce business costs in both academy and practice.