Visual Analytics - Clustering (1)

This video lecture is dedicated to clustering and visualizing clustering results. Here, we focus on global clustering approaches which are reasonable analytic techniques for rather low-dimensional data. We later discuss other clustering methods for higher dimensional data. Clustering is based on a similarity metric. We discuss some examples to show that the selection of dimensions and their weighting is not always straightforward. We compare several widely used clustering methods based on their properties and parameters. That means we do not discuss them at a technical level with respect to their actual implementation – that would be part of a data mining lecture - but at a higher level that should enable you to select an appropriate method for your specific problem. In addition to choosing an appropriate method, we want to make you aware of the parameters of the algorithm and strategies how to choose them. Therefore, we go a bit in detail with respect to properties of data that influence the choice of an algorithm and parameters, e.g., the number of outliers, varying density, and convexity. K-Means and fuzzy C-Means are discussed in detail. Chapters: 00:00 - Outline and Introduction 34:08 - Clustering Methods 47:22 - K-Means 58:00 - Fuzzy C-Means