Hierarchical data generator based on tree-structured stick breaking process for benchmarking clustering methods

A new variant of Hierarchical Cluster Analysis is gaining interest in the field Machine Learning, called Object Hierarchy. Being still at an early stage development, lack tools for systematic analysis Hierarchies inhibits further improvement this concept. In paper we address issue by proposing a generator synthetic hierarchical data that can be used benchmarking Hierarchy generation methods. The article presents thorough empirical and theoretical provides guidance on how to control its parameters. conducted experiments show usefulness capable producing wide range differently structured data. Furthermore, datasets represent most common types hierarchies are generated made available public benchmarking, along with developed (http://kio.pwr.edu.pl/?page_id=396).