Learning Undergraduate Data Science Through a Mobile Device and Full Body Movements

This study presents a mobile application (app) that facilitates undergraduate students to learn data science using their own full-body motion data. The app captures a user’s movements through the built-in camera of a mobile device and processes the images for data generation using BlazePose, an open-source computer vision model for real-time pose estimation. Students can be entirely involved in the data collection process through the app. Consequently, the motion data is contextually rich and holds personal relevance for them. This connection allows students to establish a direct relationship between their body movements and the corresponding motion data, facilitating a deeper understanding of the data. The app then takes advantage of this motion data as a data source to demonstrate various concepts and techniques in data science. As examples of the proposed learning framework, we introduce two learning modules, one focused on principal component analysis and the other on k-means clustering. To reduce learning demands, the app also provides various visual aids, such as interactive graphs and figures, that simplify the learning by visualizing the geometric interpretation of the motion data. Strategies to encompass other data science methods are also discussed for further improvement.