WaveGFormer: A wavelet-enhanced graph transformer for spatio-temporal traffic flow forecasting

Key Points

• Traffic flow forecasting accuracy significantly improved through wavelet-enhanced graph transformer utilization.
• The method reached a mean absolute error of 0.45 during validation across urban traffic datasets.
• Analysis using a wavelet-enhanced graph transformer approach led to better detection of traffic patterns over time.
• The findings suggest more accurate forecasting in real-world applications, highlighting the model's relevance in urban planning.