Current Research Directions
Much of my active research involves the advancement of fuzzy-rough set theory with a focus on real-world data mining scenarios. The mathematical properties of fuzzy-rough sets make them especially suited to processing real-valued data with inherent noise and vagueness.
Visualising Rough and Fuzzy-Rough Sets
Try out an interactive visualisation of rough sets, and see how fuzzy-rough instance selection operates
Fuzzy-Rough Methods & Data Reduction
Classical rough sets partition a discrete universe using equivalence relations. By extending this formulation through fuzzy relations, fuzzy-rough sets accommodate continuous-valued attributes without requiring arbitrary pre-discretization boundaries that often discard crucial information.
Current investigations target scalable feature selection, fuzzy-rough bireducts for simultaneous attribute and instance selection, and missing data imputation.
Fuzzy-Rough Nearest Neighbours & Classification
The intrinsic construction of upper and lower approximation operators lends itself naturally to distance- and similarity-based classification. Fuzzy-rough nearest neighbour (FRNN) techniques evaluate membership directly from local neighbourhoods.
These approaches have been successfully deployed in complex domains.
Presentations and Conference Talks
Slide decks from presentations and invited seminars delivered at international computational intelligence conferences and academic institutions.
Feature Grouping-Based Fuzzy-Rough Feature Selection
Presents a feature clustering and grouping mechanism designed to scale fuzzy-rough attribute reduction to very high-dimensional datasets while mitigating redundancy.
Download Slides (.pptx)Extending Propositional Satisfiability to Determine Minimal Reducts
Formulates the calculation of minimal fuzzy-rough reducts as a Boolean Satisfiability (SAT) problem, exploiting fast modern SAT solvers for exact minimal subset discovery.
Download Slides (.pptx)Fuzzy-Rough Instance Selection
Explores instance filtering and prototype selection using fuzzy-rough lower approximations to remove noise and accelerate nearest neighbour inference.
Download Slides (.pptx)Hybrid Fuzzy-Rough Rule Induction and Feature Selection
Introduces QuickRules, an efficient algorithm integrating fuzzy-rough dependency metrics directly into greedy modular rule induction for transparent classification.
Download Slides (.ppt)Interval-Valued Fuzzy-Rough Feature Selection with Missing Values
Extends fuzzy-rough feature evaluation to interval-valued representation for robust handling of incomplete and missing attribute values.
Download Slides (.ppt)Fuzzy-Rough Data Mining (Invited Seminar)
A comprehensive overview of foundational theory, algorithms, and applications of fuzzy-rough methods in data mining and machine learning.
Download Slides (.pptx)Professional Activities
Associate Editor
- IEEE Transactions on Fuzzy Systems (IEEE TFS)
- International Journal of Approximate Reasoning (IJAR)
Editorial Board Member
- Transactions on Rough Sets (Springer TRS)
- International Journal of Data Mining, Modelling and Management (IJDMMM)
Advisory Board Member
- The International Rough Set Society (IRSS)