The energy score gives us a general guideline for saving energy. but the R-squared values against multifamily housing and office are 0.159 and 0.025. The multivariable method that suggested in this paper gives way higher R-squared value 0.744 and 0.832. The very high value of R-squared could be a high correlation between variables. This correlation can be checked by PCA correction table and the PCA eventually choose only one principal axis. The PCA results on Multifamily Housing doesn't categorize the outcome, source EUI. However, PCA on office Building shows some clustering. If we could trial and error on PCA, office source EUI might be categorized by PCA. Proferty volum size should be a big consideration of building energy consumption because energy consumption will be affected by a triple of the building volume but the data set doesn't have direct the volume data, rather than having indirect data, numbers of floor and property floor area. If we have more data set, I would like to apply the multivariable model on historical data and find out confidence level.