Figure 4 - Linear model fit and residual plot for Office buildings. R-squared = 0.275

Random Forest Model

\label{random-forest-model}
The random forest model can explain 53.7% of the EUI variance, with an out-of-sample r-squared of 0.24 with 33% data as test data, which indicates a much better predictive power than the linear model, especially within the training set. Table 9 below shows the feature importance for each of the independent variables included in the model.
Table 9 - Independent variables for Random Forest