Dr Mykola Pechenizkiy – Eindhoven University of Technology
Predictive Analytics that Works!?
Application-driven research in predictive analytics contributes to the massive automation of the data-driven decision making and decision support. As data mining researchers and data scientists we often have a (false) believe that our techniques are immediately applicable for solving real problems, and have no bad intents; and thus we can keep our focus on developing novel techniques pushing for higher and higher accuracy of predictive models. Some of us study how to make them more robust or adaptive to changes in known and hidden contexts, others – how to facilitate privacy-preserving or privacy-aware analytics. In the first part of my talk, I will overview some of such practical issues that matter in real applications and relate them to the current state of the art in predictive analytics research.
However, recent reports as e.g. 2014 Whitehouse Review of Big Data argue that “big data technologies can cause societal harms beyond damages to privacy”, that data-driven decisions could have discriminatory effects even in the absence of discriminatory intent, that there are threats of opaque decision-making and call for a thorough studying of these threats and of methods to address them. In the second part of my talk I will revisit these concerns in the context of the personalized medicine research with the goal to highlight why the general public, domain experts or policy makers may consider predictive analytics as a thread. I will present my subjective view on what questions need to be included into the data science research agendas for gaining a deeper understanding what it means for predictive analytics to be ethics-aware and accountable and how we can achieve this.
Mykola Pechenizkiy is Associate Professor in Predictive Analytics at the Department of Computer Science, Eindhoven University of Technology (TU/e), the Netherlands. He received his PhD in Computer Science from the University of Jyvaskyla, Finland in 2005. Since June 2013 he is also Adjunct Professor in Data Mining for Industrial Applications there. His expertise and research interests are in predictive analytics and knowledge discovery from evolving data, and in their application to real-world problems in industry, commerce, medicine and education. He develops generic frameworks and effective approaches for designing adaptive, context-aware predictive analytics systems. He has actively collaborated on this with industry. He has co-authored over 100 peer-reviewed publications and co-organized several workshops, conferences, special issues, and tutorials in these areas. He served as the chair of the steering committee of Computer-Based Medical Systems (CBMS) conference series in 2012-2016. As a panelist and an invited speaker he has been advocating for the ethics-aware predictive (learning) analytics research at several recent events, including the FATML@ICML 2015 and NSF IRB Privacy and Big Data workshops and the EDM 2015 conference.
School of Computing Science & Digital Media, Robert Gordon University, Sir Ian Wood Building, Garthdee, Aberdeen, Conference Room N242, 14:00 – 15:00.