THE EFFECT OF SECURITY IN THE GREEN BUILDING PRICE PREDICTION MODEL: A COMPARISON BETWEEN MULTIPLE LINEAR REGRESSION AND MACHINE LEARNING APPROACHES

Abstract

Green building (GB) and building security are two pivotal factors that significantly influence the valuation of property prices. Nevertheless, the research on these determinants was very limited and no empirical study was done to prove the reliability of the factors as price determinants for green building. Hence, this study examines the factors by using two distinct approaches, namely the Multiple Regression Model (MRL) and Machine Learning (ML) to fill the existing empirical gap. With MRL as the conventional approach and ML as an advanced technique, the results were compared to provide maximum effectiveness in analysing the factors included. The data analysis was conducted based on a real GB dataset collected, which comprises 240 green building transactions in the city area of Kuala Lumpur, Malaysia. Prior to MLR modelling, an ANOVA test was conducted to test the statistical significance of all the independent variables (IVs) used in this study, while ML used the algorithm consisting of random forest, decision tree, linear regressor, ridge and lasso. The results indicate that building security has a strong and statistically significant impact on the price of green buildings in the MLR model. However, when it comes to enhancing prediction accuracy using the Random Forest and Decision Tree algorithms in ML models, building security has a relatively minimal influence. These results highlight a substantial difference between the outcomes of the two approaches. Specifically, the machine learning model did not demonstrate a significant relationship between green building attributes and price prediction, whereas the multiple regression model suggests otherwise.