Introducing unsupervised machine learning through a simple experiment on measuring the density of solids

Abstract

This paper proposes a pedagogical framework for introducing unsupervised machine learning techniques through a hands-on experiment conducted in a physics laboratory, focused on measuring the densities of plastic solids. The experiment involves analyzing various types of plastics by determining their densities through straightforward mass and volume measurements. The collected data are subsequently processed using clustering algorithms, including K-means, DBSCAN, and Gaussian Mixture Models, to classify the materials based on their intrinsic properties. This interdisciplinary approach bridges concepts from experimental physics and machine learning, offering students an integrative learning experience that emphasizes both data analysis and measurement methodologies. The findings highlight the effectiveness of practical applications in enhancing students’ understanding of clustering algorithms while fostering critical thinking and analytical skills in data interpretation.

Keywords
Experimental physics; machine learning; clustering