A Flexible Weighting Framework For Converting Relational Database To Hypergraphs

We propose a framework for converting relational databases (RDB) into hypergraphs to adjust and output various PageRank (PR) correlations by weighting hyperedges and nodes. While analyzing PR of converted graphs can reveal multifaceted information about relationships in the data, conventional methods struggled to balance the correlations between node PR and degree (NPR correlation) and edge PR and shared field records (EPR correlation). Our approach introduces two exponential weighting parameters: α for edge size, influencing EPR correlation, and β for node degree, influencing NPR correlation. By adjusting these parameters, various PR correlations can be obtained. Evaluation using real-world data demonstrate that our model can convert RDBs into hypergraphs with flexibility.