The UBIM approach aims at predicting long-term building performance at daily timescales using utility bills only with no monitoring at all.
The time series plots (Figure 4) illustrate the difference between the baseline energy use and data predicted from the HIM-D and UBIM modeling approaches.
The errors for HIM-D, however, are lower compared to those of UBIM. Since the CV (%) is a better indicator of the predictive accuracy when analysis is done at monthly timescales, one would unambiguously conclude that HIM-D allows for more accurate predictions on both monthly as well as annual timescales.
The average monthly NMBE errors for HIM-D range from -15% to 13%, and for UBIM from -26.8% to -9%.
Table 6 summarizes the results for the UBIM method for the service hotel located in Washington DC.