The Third European Conference on Principles of Data Mining and Knowledge Discovery

Ljupco Todorovski

The Third European Conference on Principles of Data Mining and Knowledge Discovery was held in Prague, Czech Republic, from 15th to 18th of September 1999.

76 papers were presented in the conference proceedings, edited by Jan M. Zytkow and Jan Rauch and published in the Lecture Notes in Artificial Intelligence, Number 1704, Springer. The accepted papers were divided in two categories: 28 oral presentations and 48 poster presentations. In addition to poster sessions each poster paper has been briefly presented in 3-minute highlight presentation at a plenary session.

There were also 3 invited talks. The first was on the topic of standardization of the Data Mining process by Wirth, R. from DaimlerChrysler AG, Ulm, Germany. Lehner, W. from the IBM Almaden Research Center, USA, talked on the topic Data Warehousing and OLAP Technology for Business Intelligence Applications. The last invited talk titled Bayesian hierarchical modeling, variable selection, and utility was given by Draper, D. from University of Bath, UK.

The 28 oral presentations were clustered in eight sessions. One of the sessions was titled Logic Methods (chaired by Luc De Raedt) and gathered some of the papers interesting for the ILP community.

Paper Experiments in Meta-level Learning with ILP (Todorovski, L. and Dzeroski, S.) reports on experiments in using ILP for inducing the relation between the performance of three classification algorithms to the characteristics of the dataset. In the paper Boolean Reasoning Scheme with Some Applications in Data Mining (Skowron, A. and Nguyen, H. S.) a general encoding scheme for wide class of problems (from data reduction and feature selection to pattern extraction and decision rules induction) along with efficient heuristic search for solutions, based on Boolean reasoning, were presented and empirically evaluated. Third paper in the session was On the Correspondence between Classes of Implicational and Equivalence Quantifiers (Ivanek, J.), where different implication quantifiers based on truth functions are investigated and used to construct equivalence and double implication quantifiers. The paper Querying Inductive Databases via Logic-Based User-Defined Aggregates (Giannoti, F. and Manco, G.) presents the use of logic-based database language for support of the KDD process and shows the use of the formalism for specifying rules form and evaluation function for association rules.

In the Applications session the use of ILP for predicting river water quality was presented in the paper Simultaneous Prediction of Multiple Chemical Parameters of River Water Quality with TILDE (Blockeel, H. et al.).

Last day, the Discovery Challenge was held as a part of the PKDD-99 conference. Two real-world datasets (one from the financial domain, and other from medical domain) along with brief descriptions were made available few months before paper submission deadline. 10 papers were presented, 7 on the financial dataset and 3 on the medical. One of the presented papers (Miksovsky, P. et al.) presented the use of ILP system Progol for discovering dependencies among bank branches. All papers presented in the Discovery Challenge Program are available on the DCP Web site http://lisp.vse.cz/pkdd99/Challenge/.

Additional material about the PKDD-99 conference can be found on line at: http://lisp.vse.cz/pkdd99/.