Abstract
Decision trees are a fairly mature field of research but have great potential for further research both theoretical and applied level. The proposed thesis will have as decision trees main field of research is summarized in the following stages roughly:
• Study of various classical decision tree construction algorithms, both static and dynamic (incremental).
• Study of the existing methodology construction decision tree using non-conventional methods or by applying a non-standard fields, e.g. with genetic algorithms, stochastic algorithms, hybrid techniques (reinforcement learning), in the privacy preservation problems, etc.
• Evaluation and combination of various techniques (genetic algorithms, dynamic learning, learning by resource constraints, etc.) In order to build new algorithms, the theoretical and the experimental foundation and, where possible, their application to real problems.
A very interesting topic for research is to preserve the privacy of sensitive patterns when inducing decision trees. A proposed technique is the record augmentation approach for hiding sensitive classification rules in binary datasets. Such a hiding methodology is preferred over other heuristic solutions like output perturbation or cryptographic techniques – which restrict the usability of the data – since the raw data itself is readily available for public use.
Advisory Committee
Supervisor: Dimitrios Kalles, Associate Professor, HOU
Vasilis Verykios, Professor, HOU
Christos Makris, Associate Professor, University of Patras