QuestionQ66

VerifiedModeling, Analysis, and Outcomes

A data scientist is developing a model that uses several discrete and continuous features, plus one categorical feature. What is the best reason to drop one of the resulting one-hot-encoded features when converting the categorical feature?

Explanation

For a categorical feature encoded into indicator variables, all category indicators sum to one for each record. Keeping every indicator alongside an intercept creates perfect multicollinearity; omitting one indicator establishes a reference category and removes the redundant linear relationship.

Community Discussion

No comments yet. Be the first to start the discussion!