Computer Science Seminar | Clustering: Basic Approaches and Their Evaluation
Monday, February 24, 2020
11:00 am - 12:00 pm
The widespread adoption of Internet of Things paradigm has led to an abundance of unstructured data. Accurate and timely analysis of the data is imperative to extract meaningful patterns and gain insights into the data. Consequently, several machine learning techniques have been proposed for the analysis of data, to date.
While supervised learning techniques rely on prior application knowledge and use pre-defined labels to classify data into groups, unsupervised learning techniques rely on identification of natural groupings in the data. Clustering is one of the most popular unsupervised learning techniques that uses a set of feature vectors to group data based on variability and dissimilarity metrics. This talk discusses the fundamentals of clustering and provides an overview of 3 main clustering approaches partitional, hierarchical and density-based. The benefits and limitations of each approach are discussed using a representative algorithm.
- PGH 232