Strategic Resource Trend Prediction Based on TCN-Transformer Fusion Module

This study aims to develop a high-precision battlefield resource trend prediction model to achieve comprehensive situational awareness and agile response to emergent battlefield incidents. Addressing the challenge of insufficient spatial correlation feature extraction in battlefield resource performance time series forecasting, we propose a collaborative modeling approach. This method employs a hybrid architecture of Temporal Convolutional Networks (TCN) and Transformer modules. Initially, TCN is utilized to capture short-term dependencies in temporal data, while the Transformer module effectively extracts long-term dependencies through its robust self-attention mechanism. Subsequently, a cross-attention mechanism is employed to integrate features extracted from both models, thus enhancing prediction accuracy. Experimental results demonstrate that compared to the use of TCN and Transformer individually, the combined model exhibits superior accuracy and applicability in predicting dynamic scheduling of battlefield resources and responding swiftly to emergent situations.