Prediction of building energy usage and its uncertainty analysis are critical to characterize the baseline performance of any building for impactassessment of energy saving schemes such as fault detection and diagnosis (FDD), control policies and retrofits. This paper presents a novelapproach based on Gaussian Mixture Regression (GMR) for modeling building energy use with locally adaptive uncertainty quantification. Thechoice of GMR is motivated by two key advantages (1) the number of unique building operation patterns can be identified using informationtheoreticcriteria in a data-driven manner, and (2) confidence bounds on baseline prediction are localized and their estimation is integrated with themodeling process itself. Results are presented based on synthetic data sets generated by DOE reference model for a supermarket in Chicago climateand compared with some prevalent baseline building energy models. GMR approach was found to be comparable to the polynomial model in termsof prediction accuracy of building energy consumption. However, the uncertainty in the estimated energy was found to be much more consistent withthe observed data in the case of GMR than the polynomial regression model.