Cluster Analysis

This is another chapter on multivariate analysis for classification of data, after studying logistic regression, probit models, discriminant analysis and principal components analysis (PCA). In this short chapter on cluster analysis, we begin with the concept and use cases of cluster analysis followed by a discussion of two broad approaches to cluster analysis: hierarchical and k-means, using the built-in dataset of R, mtcars. Both these approaches help classify our data, measured on ratio/interval scale, into some categories. In hierarchical clustering, the number of categories is to be determined, and it could be subjective. Though hierarchical clustering usually is the pre-step of k-means clustering. Because the hierarchical approach is resource-intensive and time-consuming, Elbow or Silhouette methods are used to determine the value of k. We discuss the Elbow method in detail and then continue with the same dataset, we explain the procedure to undertake k-means clustering.