TCPP: A Benchmark Platform for Traffic Congestion Prediction with Multi-Source Auxiliary Data

With rapid urbanization, traffic congestion has become a challenge to efficient urban transportation. To address this challenge, we develop a novel traffic congestion prediction platform (TCPP) which is a unified benchmark tool for traffic congestion prediction with multi-source auxiliary data. By combining data integration, model selection and evaluation modules, TCPP provides a standardized framework for researchers to compare and evaluate different models and datasets. Meanwhile, through a comprehensive and detailed configuration syntax, TCPP allows users to flexibly schedule tasks and execute multiple tasks continuously and uninterruptedly. In addition, two datasets collected from California highways are enriched with multi-source weather data and temporal features. We evaluate the performance of TCPP on these two datasets and demonstrate its effectiveness in predicting traffic congestion. The results demonstrate that TCPP can contribute to future research on traffic congestion prediction and the development of intelligent transportation systems and urban traffic management.