Studies in Frequent Tree Mining

Abstract

Employing Data mining techniques for structured data is particularly challenging, because it is commonly assumed that the structure of the data encodes part of its semantics. As a result are classical data mining techniques insufficient to analyze and mine these data. In this thesis we develop several mining algorithms for tree structured data and discuss some applications. Moreover, we focus on algorithms that only retrieve a small subset of all potentially interesting patterns, while the overall quality of the retrieved subset is as good as the complete set of patterns. The results show beside a smaller set of more focused patterns, that the proposed algorithms are far more efficient over existing algorithms.