Which method estimates the probability that something belongs to a particular class, and then uses a cutoff to make a yes/no decision?

Which method estimates the probability that something belongs to a particular class, and then uses a cutoff to make a yes/no decision?

• A. Linear Regression
• B. K-Nearest Neighbors
• C. Logistic Regression for Classification
• D. Support Vector Machines

Answer: (C)

Estimating the probability that something belongs to a particular class and then applying a cutoff to decide yes or no is a hallmark of logistic regression for classification. Logistic regression uses the logistic function to map input features to a probability between 0 and 1, specifically giving P(y = 1 | x). With that probability in hand, you can choose a threshold (often 0.5) to assign the binary label, or adjust the threshold to balance false positives and false negatives as needed. This combination of probabilistic output and a decision cutoff is what makes logistic regression a natural fit for this task. Other methods don’t fit as neatly. Linear regression predicts a continuous value and isn’t constrained to [0, 1], so it isn’t a reliable probability estimate for class membership. K-Nearest Neighbors can yield a local estimate of probability based on neighboring points, but that’s an empirical proportion rather than a calibrated model of the probability across the feature space. Support Vector Machines provide a decision boundary and, by default, don’t output probabilities (though they can be calibrated later); they don’t provide a direct, principled probability estimate to use with a cutoff.

Estimating the probability that something belongs to a particular class and then applying a cutoff to decide yes or no is a hallmark of logistic regression for classification. Logistic regression uses the logistic function to map input features to a probability between 0 and 1, specifically giving P(y = 1 | x). With that probability in hand, you can choose a threshold (often 0.5) to assign the binary label, or adjust the threshold to balance false positives and false negatives as needed. This combination of probabilistic output and a decision cutoff is what makes logistic regression a natural fit for this task.

Other methods don’t fit as neatly. Linear regression predicts a continuous value and isn’t constrained to [0, 1], so it isn’t a reliable probability estimate for class membership. K-Nearest Neighbors can yield a local estimate of probability based on neighboring points, but that’s an empirical proportion rather than a calibrated model of the probability across the feature space. Support Vector Machines provide a decision boundary and, by default, don’t output probabilities (though they can be calibrated later); they don’t provide a direct, principled probability estimate to use with a cutoff.