Smart Meter Anomaly Detection Dataset Pseudo Labelled via Ensembled ML Models

This dataset contains hourly electricity consumption readings from 74 smart meters, collected over a year as part of the ENFIELD smart metering infrastructure. Each record is pseudo-labeled for anomalies using a high-performing ensemble of XGBoost and LightGBM models trained on the LEAD dataset (year-long data from 200 buildings). These labels enable supervised training of interpretable anomaly detection models such as Explainable Boosting Machines (EBM).