Unsupervised Learning Methods using R

About

This five‑day advanced training course provides a comprehensive and practice‑oriented introduction to unsupervised learning methods using the R programming language. Designed for researchers, data analysts, and graduate students, the program blends conceptual foundations with extensive hands‑on sessions. Each day consists of a 3‑hour interactive module (14:00–17:00) combining lectures, demonstrations, and guided exercises in R. Participants will learn how to explore high‑dimensional data, uncover latent structures, and apply state‑of‑the‑art clustering and dimensionality‑reduction techniques to real‑world datasets.

Day 1 — Foundations of Unsupervised Learning & Introduction to R Workflows (14:00–17:00)
This session introduces the principles of unsupervised learning, including the distinction between supervised and unsupervised paradigms, the role of exploratory data analysis, and the challenges of high‑dimensional data. Participants will review essential R workflows for data pre-processing, scaling, distance metrics, and visualization. Practical exercises will focus on preparing datasets for downstream unsupervised learning tasks.

Day 2 — Dimensionality Reduction: PCA, t‑SNE, and UMAP (14:00–17:00)
Day 2 explores major techniques for reducing dimensionality while preserving structure. The module begins with Principal Component Analysis (PCA), covering variance decomposition, loadings, scores, and interpretation. It then introduces t‑SNE and UMAP, highlighting their strengths for nonlinear manifold learning and visualizing complex datasets. Hands‑on R sessions will use packages such as stats, Rtsne, and umap to generate and interpret low‑dimensional embeddings.

Day 3 — Partitioning Methods: K‑Means and Variants (14:00–17:00)
This session focuses on K‑means clustering, including algorithmic intuition, initialization strategies, cluster validation, and limitations. Extensions such as K‑medoids and silhouette analysis will also be discussed. Through guided R exercises, participants will apply K‑means to real datasets, evaluate cluster quality, and visualize cluster assignments in reduced‑dimensional spaces.

Day 4 — Hierarchical and Density‑Based Clustering (14:00–17:00)
Day 4 covers hierarchical clustering (agglomerative and divisive), linkage criteria, dendrogram interpretation, and cluster cutting strategies. The session then introduces density‑based clustering, including DBSCAN and related methods, emphasizing their ability to detect irregular cluster shapes and noise. Practical R work will use packages such as hclust, cluster, and dbscan to compare clustering behaviors across methods.

Day 5 — Model‑Based Clustering & Integrated Case Studies (14:00–17:00)
The final session introduces model‑based clustering, focusing on Gaussian mixture models, expectation–maximization (EM), and model selection using BIC. Participants will implement model‑based clustering in R using the mclust package. The course concludes with an integrated case study combining dimensionality reduction and multiple clustering techniques, enabling participants to build complete unsupervised learning pipelines from raw data to interpretable results.

Course leader

Dr. Emmanuel Abatih & Dr. Ndah Elvis

Course aim

By the end of the course, participants will be able to:

Understand the theoretical foundations of major unsupervised learning methods.

Apply PCA, t‑SNE, and UMAP for dimensionality reduction and visualization.

Implement K‑means, hierarchical, density‑based, and model‑based clustering in R.

Evaluate and compare clustering solutions using quantitative and graphical tools.

Construct reproducible unsupervised learning workflows for research and applied projects.

Fee info