Hourly Anomaly Scores and Leak Labels from a Multi-Source Urban Water Distribution Network Dataset

Dataset: Hourly Anomaly Scores and Leak Labels from a Multi-Source Urban Water Distribution Network Dataset

Description:
This dataset includes hourly time series from an anonymized urban Water Distribution Network (WDN), combining SCADA sensor readings, operational energy usage, and environmental conditions. Anomaly scores were generated using Elastic ML and Isolation Forest. Binary labels indicate proximity (±7 days) to real leak events.

Columns:
- timestamp: DD-MM-YYYY HH:00
- fault_d7: Binary leak proximity label
- *_kW, *_power_hour: Energy-based anomaly scores
- *_ws_*: Sensor-based anomaly scores (pressure, temp, vigor)
- temp_site1_anomaly_score: Environmental temperature anomalies
- gw_lvl_site2_anomaly_score: Groundwater level anomalies