From the course: Applied AI Auditing in Python

Unlock the full course today

Join today to access over 24,700 courses taught by industry experts.

Error analysis

Error analysis

- [Instructor] An error analysis is a valuable tool in the AI auditing process, particularly when access to the underlying model is limited or unavailable. In this section, instead of building a model, you'll be provided with a dataset of predictions. This will simulate an external audit where you only have access to outcomes. In this video, I'll walk you through an error analysis before you get to conduct one on your own. In CoLab, we want to make sure we have the Error Analysis IPython Notebook open, and if we click on this folder in the left-hand side, you can easily drag over our defaulted loans CSV. Once that's uploaded, we can go ahead and install Fairlearn. Once that's done, we can go ahead and get our helper packages installed. And now let's go ahead and import our dataset. Now we can explore the dataset by displaying the first few rows using the head function. You can see this dataset has three columns, our actual defaults, gender, and the probability of default. Let's print…

Contents