Material Flow Prediction Task Based On TCN-GRU Deep Fusion Model

With the rapid growth of e-commerce and smart logistics, accurate material flow prediction has become essential for improving delivery efficiency, reducing operational costs, and supporting decisions such as warehouse scheduling and route optimization. This study proposes a material flow forecasting method based on a deep fusion model combining Temporal Convolutional Networks (TCN) and Gated Recurrent Units (GRU), designed to enhance prediction accuracy and capture both short- and long-term temporal dependencies. The model is evaluated using the Amazon Delivery Dataset, which includes 43,632 delivery records from multiple cities, containing detailed information on orders, agents, traffic, weather, and performance metrics. Focusing on the city of Chennai, India, the study further explores how weather conditions influence delivery outcomes. The TCN module effectively extracts short-term temporal features, while the GRU component captures longer-term trends. Experimental results show that the proposed TCN-GRU model achieves a Mean Absolute Error (MAE) of 1.860, outperforming traditional forecasting models in both accuracy and generalization. These findings highlight the model’s potential for real-world logistics applications and contribute to advancing intelligent logistics systems under complex, dynamic conditions.