Volodymyr Melnykov
January 30th, 2020 3:00-4:30 pm @ Gorgas Library 104
The Power of Model-based Cluster Analysis
Cluster analysis is an important area of unsupervised machine learning with numerous applications in all fields of human activity. The abundance of existing cluster analysis methods oftentimes are used rather arbitrarily by practitioners. However, such application of clustering procedures is likely to lead to misleading or erroneous results.
Model-based clustering relying on the notion of finite mixture models is one of the most flexible and powerful tools for partitioning data. It assumes that each data group can be seen as a sample from a corresponding mixture component. The one-to-one relationship established between data groups and mixture components makes model-based cluster analysis highly interpretable.
An introduction to cluster analysis will be provided with a specific focus on model-based methods. Main advantages and challenges related to the approach will be discussed along with the most useful software tools. A variety of extensions and applications will be considered.
Feel free to bring your laptop and install both R and Rstudio (https://www.andrewheiss.com/blog/2012/04/17/install-r-rstudio-r-commander-windows-osx/). Gorgas 104 has computers with R ready to be used as well.
About the Instructor
Dr. Volodymyr Melnykov is a Professor of Statistics in the department of Information Systems, Statistics, and Management Science and an Elected Director of the Classification Society of North America. He received his PhD at Iowa State University in 2009 and published over 30 papers devoted to the theory and applications of cluster analysis.
