A tutorial on unsupervised Gaussian mixture model for performance clustering in second language research

This tutorial introduces the application of unsupervised Gaussian Mixture Model (GMM) clustering to identify second language (L2) performance profiles. GMM employs a probabilistic clustering technique that accommodates overlapping profile membership and provides a flexible method for analyzing high-dimensional performance data commonly encountered in L2 research. Using L2 writing assessment data from a local English placement test, we present a step-by-step analytical pipeline, covering data preparation, dimensionality reduction, model selection, visualization, and interpretation. This approach is adaptable to other performance modalities (e.g, speaking) and can be enriched with additional performance features to support a more comprehensive understanding of L2 performance and underlying language ability.