The growing complexity of urban traffic networks necessitates advancements in traffic flow prediction. Traditional methods frequently fail to deliver accurate predictions, primarily due to their oversimplified nature and the omission of key factors, notably weather conditions. To overcome these deficiencies, this paper proposes the Weather-Influenced Attention-Based Spatio-Temporal Graph Convolutional Network (WI-ASTGCN), an advanced refinement of the ASTGCN. This model judiciously incorporates weather weighting factors, chosen for their strong correlation with traffic flow, such as atmospheric temperature, rel- ative humidity, and wind speed. It adeptly captures the complex spatiotemporal dynamics and adapts in real-time to weather changes, enhancing prediction accuracy. The experimental results demonstrate that the model has better predictive performance compared to the six existing baseline methods. Compared to the ASTGCN model, the overall performance of the model in terms of reduction of prediction error is improved by about 8 \(\%\) .