AI Engineering Degree Practice Exam

Logistic Regression is designed specifically to handle categorical outcomes, making it analogous to linear regression but tailored for such targets. While linear regression predicts a continuous dependent variable, logistic regression predicts probabilities that map to categorical outcomes, typically binary (such as 0 or 1, yes or no). One of the key aspects of logistic regression is the use of a logistic function (sigmoid) to model the probability of a particular class. This transformation allows it to effectively handle situations where the dependent variable is categorical, thereby distinguishing it from linear regression, which isn't suitable for this type of data. Other statements, such as the requirement for numeric targets or the need for continuous outcomes, are inaccurate because logistic regression does not rely on continuous targets for its predictions. Additionally, logistic regression can be extended to multi-class classification through techniques such as One-vs-Rest (OvR) or Softmax regression, thus it isn't limited to binary classification alone. This flexibility reinforces the validity of the correct statement regarding its applicability to categorical targets.