Research · 2026
ConfoSense.
Making building alerts more reliable by giving forecasts a measure of uncertainty. A common protocol connects forecasting, prediction intervals and operational alerting.
The problem
Accurate forecasts alone do not tell a building operator which deviations deserve attention.
My contribution
I developed a common protocol connecting forecasting, conformal prediction and causal alert generation. It compares temperature, energy and HVAC tasks using detection, alert workload, delay and robustness measures.
The outcome
The public results index records the evaluation scope and supporting reports. The work investigates how uncertainty calibration changes operational alerting decisions, rather than claiming a new forecasting or conformal algorithm.
Scope & status
Results are bounded to the declared datasets and synthetic-fault protocols. This is research evidence, not a deployed building-control system.