Which method is used to predict probabilities of outcomes rather than final class label?

• A. Linear Regression
• B. Logistic Regression
• C. Discriminant Analysis
• D. Naive Bayes

Answer: (B)

Predicting probabilities for outcomes relies on a model that outputs P(Y|X) directly. Logistic regression does exactly that: it combines the input features linearly and then applies the logistic function to squash the result into the interval [0, 1], giving a probability for the positive class for each observation. This probabilistic score is what you use when you need a measure of likelihood, not just a hard decision, and you can work with it to set thresholds or compute metrics like ROC-AUC.

Linear regression isn’t designed to produce probabilities; it predicts a continuous value and can yield numbers outside [0, 1], which aren’t interpretable as probabilities without extra steps. Discriminant analysis and Naive Bayes can provide posterior probabilities as part of their outputs, but they rely on distributional assumptions about the data. Logistic regression is the most straightforward and widely used approach when the goal is directly estimating the probability of a binary outcome (and can be extended to multiple classes with methods like multinomial logistic regression).