What step is NOT part of framing a typical machine learning (ML) problem?

Framing a typical machine learning problem involves several key steps that help set the foundation for successful model training and evaluation. The first three choices—defining model evaluation criteria, preparing the dataset for training, and determining the initial feature set—are all integral components of the problem framing process.

Defining model evaluation criteria is essential as it establishes the metrics by which a model's performance will be judged. This ensures you have a clear understanding of what success looks like for your machine learning endeavor.

Preparing the dataset for training is crucial too. This step typically includes cleaning, processing, and transforming raw data into a format suitable for input into a machine learning model, which is fundamental for achieving accurate results.

Determining the initial feature set involves selecting the relevant variables that will be used as inputs for the model. This selection is critical because the right features can significantly impact model performance.

In contrast, running an initial test set of 20 percent of the data to establish the appropriate schema does not fit within the framing step. While testing and validating the model's performance on a holdout dataset is an important aspect of the overall machine learning workflow, it occurs after the problem has been framed and the model trained. Therefore, this action is more related to model evaluation rather