Multi Source Heterogeneous Correlation Data Fusion Based on Graph Clustering and DFS Algorithm

In multi-source heterogeneous correlated data, the correlation between data often needs to be established through multiple intermediate nodes or paths. This multi hop correlation may be lost during the data fusion process due to algorithm limitations, resulting in incomplete correlation mining and affecting the fusion effect. Therefore, a multi-source heterogeneous correlation data fusion method based on graph clustering and DFS algorithm is proposed. Assign unique identifiers to entities and map them to graph nodes, construct a set of nodes and edges, and ultimately form a graph structure represented by an adjacency matrix. Using this as input to the DFS algorithm, capturing the multi hop correlations between data by recording the weights and correlations of multi hop paths, and updating the correlation matrix to reflect these correlations. At this point, the spectral clustering method is used to embed nodes into a low dimensional space using the eigenvalue decomposition of the Laplacian matrix. Then, the K-Means algorithm is combined to cluster the nodes, effectively capturing the global structural information of the data and achieving efficient fusion of multi-source heterogeneous correlated data. The experimental results show that when the data volume reaches 500 GB, the multi hop correlation coverage of the proposed method remains at 97.4 %, and the information loss rate is only 0.12 %. It can effectively integrate multi-source heterogeneous correlated data fusion and improve the effectiveness of data.