This study applies graph-based machine learning techniques to model and analyze bibliographic networks, focusing on co-citation and co-author relationships. By transforming bibliographic datasets into graph structures representing authors, papers, and citations, the research leverages machine learning algorithms such as Fast Random Projection for graph embedding and Logistic Regression for link prediction and node similarity. This study also evaluates challenges in Author Name Disambiguation (AND). The study demonstrates the advantages of graph databases in handling complex, interconnected scholarly data, offering superior performance compared to traditional databases. The findings provide valuable insights for fields like bibliometrics and Science of Science (SciSci) and highlight the potential for future research in large-scale bibliographic data analysis using graph databases.