Research on the correlation analysis of investigation clues and evidence chain reconstruction model based on multi-source data fusion

To address the high discretization of multi-source heterogeneous data and the difficulty of reconstructing logically complete evidence chains in criminal investigations, this paper proposes a multi-source data fusion method for investigative clue association analysis. First, a spatiotemporal feature encoder maps text, image, and structured data into a unified semantic space. Second, a heterogeneous graph attention network is constructed with case entities as nodes to automatically discover deep associations among personnel, vehicles, communication devices, financial accounts, and locations. Third, a legal knowledge graph is introduced for logical pruning and consistency verification, transforming data-driven clue associations into legally interpretable evidence paths. Finally, a visualized closed-loop evidence-chain topology is generated to support judicial interpretability and investigator review. Experimental results show that the proposed model achieves 92.5% precision, 89.3% recall, a 93.5% implicit-connection detection rate, and an evidence-chain logical integrity score of 94.2, representing significant improvement over comparative methods. The model provides a technical framework for intelligent investigation, multimodal evidence fusion, and explainable evidence-chain reconstruction.