PhD defence
Exploring graph-based clustering and outlier detection algorithms
- J. Li
- Date
- Wednesday 12 November 2025
- Time
- Location
-
Academy Building
Rapenburg 73
2311 GJ Leiden
Supervisor(s)
Summary
In the big data era, traditional clustering and outlier detection methods often fail to capture complex relationships in high-dimensional, noisy, or dynamic data due to reliance on Euclidean distances or global thresholds. Graph-based approaches, leveraging structural properties, offer superior adaptability and efficiency.
Minimum spanning tree (MST)-based clustering detects irregularly shaped clusters and hierarchical structures by analyzing graph Laplacians, outperforming density-based methods like DBSCAN in sparse data. For outlier detection, Spectral Analysis Outlier Detection (SAOD) identifies anomalies in high-dimensional spaces through eigenvalue distributions of kNN graphs, excelling in medical diagnosis and cybersecurity. Adaptive Mini-MST Outlier Detection (MMOD) overcomes density dependency using scaled distance metrics, while Scaled MST with Medoid Selection (MS2OD) reduces noise sensitivity via robust medoid selection.
This research establishes graph-based techniques as indispensable for modern data analysis, bridging theory and practice. Future work will focus on real-time streaming data analytics and interpretable frameworks to enhance trust and applicability.
PhD dissertations
Approximately one week after the defence, PhD dissertations by Leiden PhD students are available digitally through the Leiden Repository, that offers free access to these PhD dissertations. Please note that in some cases a dissertation may be under embargo temporarily and access to its full-text version will only be granted later.
Press enquiries (journalists only)
+31 (0)71 527 1521
nieuws@leidenuniv.nl
General information
Beadle's Office
pedel@bb.leidenuniv.nl
+31 71 527 7211