Data are not only ubiquitous in society, but are increasingly complex both in size and dimensionality. Dimension reduction offers researchers and scholars the ability to make such complex, high dimensional data spaces simpler and more manageable. This Element offers readers a suite of modern unsupervised dimension reduction techniques along with hundreds of lines of R code, to efficiently represent the original high dimensional data space in a simplified, lower dimensional subspace. Launching from the earliest dimension reduction technique principal components analysis and using real social science data, I introduce and walk readers through application of the following techniques: locally linear embedding, t-distributed stochastic neighbor embedding (t-SNE), uniform manifold approximation and projection, self-organizing maps, and deep autoencoders. The result is a well-stocked toolbox of unsupervised algorithms for tackling the complexities of high dimensional data so common in modern
In the age of data-driven problem-solving, applying sophisticated computational tools for explaining substantive phenomena is a valuable skill. Yet, application of methods assumes an understanding of the data, structure, and patterns that influence the broader research program. This Element offers researchers and teachers an introduction to clustering, which is a prominent class of unsupervised machine learning for exploring and understanding latent, non-random structure in data. A suite of widely used clustering techniques is covered in this Element, in addition to R code and real data to facilitate interaction with the concepts. Upon setting the stage for clustering, the following algorithms are detailed: agglomerative hierarchical clustering, k-means clustering, Gaussian mixture models, and at a higher-level, fuzzy C-means clustering, DBSCAN, and partitioning around medoids (k-medoids) clustering.
