"PEMS2023-SANJOSE: A Curated Multi-Sensor Traffic Flow Dataset for Spatio-Temporal Forecasting"

"This dataset provides a cleaned and research-ready traffic flow time series derived from raw records downloaded from the Caltrans Performance Measurement System (PeMS). The raw PeMS files were processed to retain a compact and consistent representation of traffic activity across 319 monitoring stations. Each record contains three fields: timestamp, station ID, and total traffic flow aggregated over a 5-minute interval.The dataset is intended to support research on short-term traffic forecasting, spatio-temporal prediction, intelligent transportation systems, and data-driven mobility analysis. By standardizing the raw PeMS records into a simplified station-level time series format, the dataset reduces preprocessing effort and enables direct use in forecasting models, benchmarking studies, and reproducibility-oriented experiments.The cleaned structure preserves the essential temporal and sensor-specific characteristics required for traffic flow modeling while removing unnecessary raw-file complexity. This makes the dataset suitable for machine learning, deep learning, graph-based forecasting, and lightweight time-series prediction studies. Users should cite the associated article and this IEEE DataPort dataset record when using the dataset in academic or applied research."