Plenary Speakers

Prof. Katarzyna Stąpor -Silesian Technical University, Gliwice, Poland
  • Evaluating and comparing classifiers: review, some recommendations and limitations

Performance evaluation of supervised classification learning method related to its prediction ability on independent data is very important in machine learning. It is also almost unthinkable to carry out any research work without the comparison of the new proposed algorithm with other already existing methods.
This paper aims to review the most important aspects of the evaluation process of supervised classification algorithms as well as of their comparison in the following situations: 1) comparison of two algorithms on one dataset, 2) on multiple datasets, and 3) comparison of multiple algorithms on multiple datasets.
Critical view, recommendations and limitations of the reviewed methods in different supervised classification problem scenarios are presented.
Thus, we provide a quick guide to understand the complexity of the evaluation process and try to warn the reader about the wrong habits.

Katarzyna Stąpor is a Professor at the Department of Automatic Control, Electronics and Computer Science in the Silesian Technical University in Gliwice, Poland. She has published over 80-ty publications including one book on pattern recognition (“Pattern Classification Methods in Computer Vision”, PWN, Warszawa, 2011, in Polish) and the handbook “Statistical methods for students of computer science with examples in R” (Silesian Technical University Press, Gliwice, 2015, in Polish). Her current research interests include statistical pattern recognition, multivariate statistical analysis, computer vision and bioinformatics. She is also the co-author of the integrated
information management systems and the classification system supporting glaucoma diagnosis in ophthalmology. Prof. Stąpor is the member of Advisory Board of the Machine Graphics&Vision journal and the member of Editorial Board of Bio-Algorithms and Med-Systems.
Dr. Bartosz Krawczyk - Virginia Commonwealth University, U.S.A.
  • Challenges in Drifting Data Stream Mining.

Developing efficient classifiers that are able to cope with big and streaming data, especially with the presence of the so-called concept drift, is currently one of the primary directions among the machine learning community. This presentation will highlight the contemporary challenges in data stream mining such as limited access to true class labels, changing number and structure of classes and evolving imbalanced data. The talk will start with the active learning approaches that allow us to select only most important instances for label query and at the same time take into account the drifting nature of the stream. Then, rule-based and ensemble classifiers that can accommodate emergence of new classes and disappearance of old ones will be discussed. Finally, algorithms for handling binary and multi-class skewed data streams with changing imbalance ratios will be presented. The talk will highlight the fact that these challenges are not independent and we must take all of them into consideration when designing efficient online classifiers for data streams.

Bartosz Krawczyk is an assistant professor in the Department of Computer Science, Virginia Commonwealth University, USA, where he heads the Machine Learning and Stream Mining Lab. His research is focused on machine learning, ensemble learning, data streams, class imbalance, one-class classifiers, and interdisciplinary applications of these methods. He has authored 35+ international journal papers and 80+ contributions to conferences. Dr Krawczyk was awarded with numerous prestigious awards for his scientific achievements like IEEE Outstanding Leadership among others. He served as a Guest Editor in four journal special issues and as a chair of ten special
session and workshops. He is a member of Program Committee for over 40 international conferences and a reviewer for 30 journals.

Layout by Maciej Krysmann