Testing the use of machine learning for heritage property valuation / Junainah Mohamad, Nur Shahirah Ja’afar and Suriatini Ismail

Recently, the use of machine learning is gaining ground and it holds great promise for real estate valuation. However, the application of machine learning in heritage property valuation has limited adoption. Therefore, this paper aims to demonstrate the potential use of machine learning in heritage property valuation. The original dataset consists of 311 prewar shophouses transacted from 2004 to 2018 at North-East of Penang Island, Malaysia. After the filtering process, only 137 units of pre war shophouse heritage property were available and valid to be used. Several machine learning algorithms have been developed and tested, including random forest regressor, decision tree regressor, lasso, ridge and, linear regression. The results indicate that random forest regressor is the best machine learning algorithms and can be used for heritage property valuation.