Unsupervised Learning (Principal Component Analysis, K-Means Clustering)

Unsupervised learning is an important area of data science that focuses on identifying hidden patterns, structures, and relationships in datasets without using predefined outcome labels. This topic introduces students to the fundamental principles of unsupervised learning and its role in exploring complex datasets, discovering groups of similar observations, and reducing data complexity. Students will learn how unsupervised learning differs from supervised learning and understand its applications in fields such as public health, environmental science, economics, and social research. Practical examples, such as grouping patients based on health characteristics, identifying regions with similar environmental conditions, and exploring patterns in consumer behavior, will help students understand how unsupervised learning supports exploratory data analysis and data-driven decision-making.

The material focuses on two important unsupervised learning techniques: Principal Component Analysis (PCA) and K-Means Clustering. PCA is a dimensionality reduction technique that transforms a set of potentially correlated variables into a smaller number of uncorrelated variables called principal components, which capture as much of the original data variation as possible. Students will learn about data standardization, covariance and correlation matrices, eigenvalues, eigenvectors, explained variance, and the use of scree plots to determine how many principal components to retain. K-Means Clustering, meanwhile, is a partitioning method that groups observations into a predefined number of clusters based on their similarity, typically by minimizing within-cluster variation. Students will explore the concepts of centroids, distance measures, cluster assignment, iterative updating, and the selection of the number of clusters using methods such as the elbow method and silhouette analysis. They will also examine how scaling, outliers, and the choice of cluster number can influence the results of both techniques.

Furthermore, students will learn to implement and interpret PCA and K-Means Clustering using R and Python. Practical exercises will cover data preparation, handling missing values, standardizing numerical variables, applying PCA, visualizing observations in reduced-dimensional spaces, and identifying clusters through graphical representations. Students will learn to interpret explained variance and principal component loadings in PCA, as well as cluster centroids and cluster profiles in K-Means Clustering. They will also evaluate the strengths and limitations of these methods, including the sensitivity of K-Means to initialization and outliers, the assumption of approximately spherical clusters, and the fact that PCA prioritizes variation rather than class separation. By the end of this topic, students are expected to apply PCA and K-Means to explore high-dimensional data, identify meaningful patterns, and communicate their findings effectively in practical data science applications.