The reasoning for this approach has been provided in screen 3 of this project:
Let’s start with some univariate k-nearest neighbors models. Starting with simple models before moving to more complex models helps us structure your code workflow and understand the features better.