Traffic Prediction via Signal Decomposition and Attention-based BiLSTM


Peking University & Kunshan Duke University | Advisor: Prof. Peng Sun

February 2025 – May 2025 | Shenzhen, CN

  • Addressed challenges in capturing trend and seasonal components in traffic flow data by integrating signal decomposition with deep learning frameworks. Developed a dual-stream BiLSTM model incorporating moving average, Fourier transform, and multi-head attention mechanisms, achieving state-of-the-art performance on PeMS datasets.
  • Led the project lifecycle, including research design, algorithm development, experimentation, and manuscript preparation.
  • Submitted to IEEE GLOBECOM 2025.