Within the data mining community, several languages have been proposed to express data mining tasks and many proposals for tightly integrating data mining and databases into one "inductive database" were made. On top of that, all data mining suites offer the possibility to define a workflow to combine and manage complex compositions of basic data mining and data manipulation tasks.

As such, much work went into structuring the data mining process. Little work, however, went into the question what can be done with the results of the data mining. However, especially in unsupervised settings, such as pattern mining and clustering, user interaction is required to inspect and analyze the most useful results.

The goal of this topic is therefore to develop OLAP-alike functionality to support the evaluation of the results of data mining. Typical operations would include grouping patterns; e.g., based upon which characteristics of the data they address and summarizing several numerical properties of the patterns, such as frequency and discriminative power. The major challenges of this project are establishing a conceptual equivalent of the data cube for patterns, the identification of a meaningfull and expressive set of operators, and an efficient evaluation strategy.

Main Advisor at Université Libre de Bruxelles (ULB)
Co-advisor at Universitat Politècnica de Catalunya (UPC)