Abstract: HPC can significantly improve many data mining and business intelligence tasks, but in fact HPC is still almost not supported by leading data mining tools. Two major reasons for that are lack of efficiently paralleled software packages and extra complexity related to managing of remote HPC systems. Therefore we developed our own software tool called GMDH Shell that shows how both problems can be solved efficiently for popular data mining tasks such as classification, regression and time series forecasting.
Bio: Oleksiy Koshulko earned his doctoral degree in computer science specializing in mathematics modeling and numeral methods from Glushkov Institute of Cybernetics in 2007. Prior to this he received a Specialist degree in applied mathematics from Shevchenko Kyiv National University, Ukraine. His research interests include parallel processing in the Group Method of Data Handling (GMDH) and its parallelization and adaptation on cost-effective multiprocessor systems. Dr. Koshulko is a recipient of 2004 President of Ukraine Fellowship for Young Scientists and 2009 Award of Ukrainian Parliament for most talented young scientists. He is an IEEE member and Top500 consultant.