"Simulated Urban Traffic Anomaly Dataset: eDPF Pre-filtered and Unfiltered Data for Root Cause Localization"

"This dataset offers a valuable resource for advancing root cause localization in urban traffic anomaly analysis. Generated via an automated SUMO simulation framework, it features high-fidelity, simulated urban traffic anomaly scenarios with precise ground truth. The dataset contains two key components: raw, unfiltered time-series sensor data (flow, occupancy, speed) collected during anomalous events, and a corresponding filtered subset processed by the novel Efficient Detector Pre-Filtering (eDPF) framework. The eDPF filtering employs phase-aware robust statistics and a multi-stage aggregation process to identify detectors with elevated anomalousness scores, thereby concentrating on anomaly-relevant signals and significantly reducing data dimensionality. This curated collection supports rigorous development, evaluation, and comparison of root cause localization algorithms, enabling researchers to assess the impact of pre-filtering on computational efficiency and localization accuracy within complex urban traffic networks."