Occupancy in the field of HVAC design is crudely estimated early in the design phase and often does not reflect the actual building. This results in chronic overventilation, which wastes significant energy as conditioning outdoor air is the driver of HVAC energy use. Therefore, reducing outdoor air damper positions based on the predicted occupancy of buildings can generate significant savings. This research develops occupancy-based predictive controls for outdoor air dampers in a case study building in Ottawa, Canada. It was discovered that multiple linear regression models using Wi-Fi and electrical load data produced accurate estimates. A novel forecasting framework consisting of k-means clustering, motif identification, and classification trees was employed to develop a rules-based method for occupancy prediction that could be practically applied in existing commercial buildings. A six-month long implementation of these occupancy-based controls showed that heating and cooling energy use were reduced by 38% and 10%, respectively.