- Machine Learning (ML) suites are readily available for practical use. There are available packaged software and console libraries that can be modified. However, novice programmers avoid these ML suites due to the abstractions with pre-existing tools. Users with limited programming backgrounds that know how these models work struggle with converting them to code for practical use. We iterated on developing a code block tool for linear regression tasks where we did multiple usability tests involving at least 33 participants. We observed how they went through their ML tasks and inquired into their pains and struggles in writing code. To help them in this learning process, we present TREX: A Toolbox for Regression Experiments, which allows novices in ML programming to gain more confidence with building code. We also derived and verified design guidelines towards providing affordances that cater to their needs.
More details can be found in the project website.
Published in: CHIuXiD 19