Yesterday we covered numeric and general categorical data. Today we explore feature engineering for two specialized data types—dates and times (time series) and text—that require custom handling. Finally, we'll address the perpetual challenge in encoding: managing high-cardinality categorical features through dimensionality reduction and hashing techniques.
This topic appears frequently in MLS-C01 exams in the form "which transformation is appropriate for what data?"
Using a date column as-is (e.g., 2026-06-26 14:30:00) leaves the model unable to extract meaning. The date must be decomposed into meaningful components.