Multi-source Information Data Fusion Method between Heterogeneous Platforms

This study strives to tackle the complexities of merging multi-source data on diverse platforms and introduces a novel data fusion approach leveraging deep learning (DL). Data from various sources is seamlessly amalgamated via preprocessing techniques including data purification, format conversion, and normalization. The study employs an attention-based fusion mechanism that adaptively learns and assigns importance to features from each data source. Experimental results indicate that our method outperforms alternative approaches in terms of precision, F1 score, and other metrics, demonstrating excellent performance and stability. Notably, our accuracy and F1 score exceed $90 \%$, with a mean squared error (MSE) of approximately 4.11. This innovative approach paves the way for multi-source data fusion on heterogeneous platforms and holds promise for broad application across multiple domains. Future directions encompass assessing the method’s real-world performance, bolstering data security, seeking efficient model training techniques, and broadening its applicability.