Optimizing Traffic Flow Prediction with BiLSTM-TCN Fusion

Abstract

Accurate traffic flow prediction is essential for effective transportation management and urban planning. However, existing methods often struggle to simultaneously capture short-term fluctuations and long-term trends. To address this challenge, we propose a hybrid approach that combines Bidirectional Long Short-Term Memory (BiLSTM) and Temporal Convolutional Network (TCN) architectures. Our method involves meticulous data preprocessing and integrates BiLSTM to capture long-term dependencies and TCN to model short-term dynamics. Experimental results on the Kaggle’s traffic prediction dataset demonstrate that BiLSTM-TCN outperforms existing methods in terms of both training and validation loss. Compared to the current best-performing algorithm, BiLSTM-TCN achieves a reduction in mean training loss of 0.02 and a reduction in mean validation loss of 0.02. This corresponds to a percentage reduction of 19.05% in mean training loss and 16.00% in mean validation loss, underscoring the superior performance of BiLSTM-TCN. Additionally, BiLSTM-TCN shows significant improvements inRMSE, MAPE, and MAE, making it a valuable tool for transportation management and urban planning initiatives.