Estimates in predictive modeling can relate to probabilities for cases with what type of targets?

In predictive modeling, estimates relating to probabilities are specifically applicable to categorical targets. Categorical targets are those that can take on a limited and usually fixed number of possible values, often representing distinct categories or classes. In the context of predictive modeling, when dealing with categorical outcomes, the model predicts the likelihood or probability of each class for a given input.

In contrast, continuous targets represent a range of values and are typically associated with regression analysis, where predictions involve estimating a numeric outcome rather than probabilities. Binary targets, while a subset of categorical targets, specifically refer to those with only two categories (e.g., yes/no or true/false), and are indeed typically associated with probability estimates; however, the broader category of categorical encompasses a wider range of situations, including multi-class classifications.

Numerical targets refer generally to any continuous numeric outcome. Therefore, when discussing probabilities in predictive modeling, the appropriate and comprehensive choice is categorical targets, as they cover both binary and multi-class scenarios, as well as the fundamental nature of the probabilities associated with classifications.