PhD Defense: Supporting Independent Learning and Rapid Experimentation in Data Science

Data Science tutorials built using computational notebooks enable the audience to discover and explore new material. However, while templates and tutorials remain static---best practices, libraries, and versions evolve. Keeping up with these visualization and data analysis trends is becoming increasingly complex, especially for novice data scientists. We can automatically track analysis iterations, make analytical practice automatically comprehensible to notebook readers, and use this data to create dynamic and relevant tutorials for novice data scientists. In this thesis, we develop a novel design for a computational notebook learning environment for novices. Specifically, we model the analytical practice in notebooks authored by experienced data scientists and use this model to generate labeled and documented data analysis recommendations.

Examining Committee

Chair:

Dr. Niklas Elmqvist

Dean's Representative:

Dr. Philip Resnik

Members:

Dr. Amol Deshpande

Dr. Michelle Mazurek

Dr. Huaishu Peng

Dr. Leilani Battle