Validation-Selected Hybrid Occupancy Forecasting for Empty-Window Recommendations: An Offline Post-Bin Case Study
The Storm King School; Department of Electrical Engineering, Columbia University
2026 · Unpublished author manuscript. Submitted to ACM BuildSys; decision pending. Submission status reported by Junyu Chen on October 30, 2026.
Abstract
Building operations need occupancy forecasts that arrive in time, last long enough, and rarely conflict with later occupancy observations. We study this problem in a single-building, 15-min, offline post-bin case study. The primary hybrid is a weighted average of Empty-class scores from a historical schedule, LightGBM, and a compact Transformer, combining recurring, nonlinear, and sequential patterns. Validation area under the precision–recall curve (AUPRC) selects the weights. AUPRC summarizes Empty-class precision and recall across score thresholds; higher values indicate better discrimination when class frequencies are unequal. A separate validation search selects the operating threshold that maximizes a load-opportunity proxy under a 10% conflict-rate limit and a one-hour minimum window. Conflict rate is the share of recommended intervals later labelled occupied. The opportunity proxy sums processed HVAC-plus-lighting load in recommendations later labelled Empty and represents load available for further operational investigation. A stored anchor t covers the completed input bin [t, t + 15 min), making the forecast available at t + 15 min. Across 388,032 overlapping held-out forecast rows, hybrid AUPRC is 0.8514, compared with 0.8497 for the historical baseline and 0.8382 for LightGBM. Across 43 non-overlapping policy horizons, the fixed hybrid policy recommends 259 intervals; all are subsequently camera-label-empty, form 14 label-safe windows, and coincide with 490.1 kWh of processed load proxy. LightGBM finds a similar opportunity (493.9 kWh) with 11 conflicts, whereas the conflict-free historical policy finds 94.6 kWh. The hybrid therefore preserves strong forecast discrimination while improving the observed opportunity–conflict trade-off. A prospective rerun with verified source timestamps is the next step toward operational evaluation.
Author-written abstract, reproduced from the supplied manuscript. The complete PDF remains unchanged.
Scope and interpretation
The findings concern saved outputs from one building. The hybrid's small AUPRC advantage does not establish statistical superiority: paired uncertainty intervals cross zero. Recorded camera-empty labels do not verify physical absence, and load overlap does not establish energy savings or controllable capacity.
Source timestamp and imputation provenance, corrected model initialization, and evaluation on fresh data remain necessary for a prospective study. The software release and public manuscript do not establish peer-review acceptance.
Reproduce the fixed comparison
Download the offline Python replay kit, try the browser demo, or open the notebook. The kit recomputes the three reported policy rows using included compact records and Python's standard library.
Suggested citation
Chen, Junyu, and Chong Li. 2026. Validation-Selected Hybrid Occupancy Forecasting for Empty-Window Recommendations: An Offline Post-Bin Case Study. Unpublished author manuscript, research-record-v1.