TEACHING AND LEARNING DATA-DRIVEN MACHINE LEARNING WITH EDUCATIONALLY DESIGNED JUPYTER NOTEBOOKS

Abstract

This study examined upper secondary students’ modeling with machine learning algorithms. In the context of a yearlong data science course, the study explored how students applied machine learning using Jupyter Notebook and how they documented the modeling process as a computational essay taking into account the different steps of the CRISP-DM cycle. The students’ work was based on a teaching module about automatically created decision trees as a machine learning method. Worked examples were used to support students’ modeling processes. The study investigated the students’ performance in technically carrying out the machine learning methods and their reasoning about different data preparation steps, bias in the data, the application context, and the resulting decision model.