Symbolic Data Analysis: Adapted clustering methods for data represented with discrete distributions

Classical clustering approaches are based on data descriptions with classical vectors. Due to privacy policy and/or large datasets many data are actually aggregated and represented by discrete distributions. For them classical representations do not preserve all the information. Therefore, we adapted some well-known clustering methods and also developed some new approaches for clustering data represented with discrete distributions. These approaches and the results of their applications on real data sets will be presented in the seminar. At the end, we will also discuss some other approaches for data analysis in symbolic data analysis that we are currently working on.