Search

We study the problem where a decision maker uses a linear classifier over attribute values (e.g., age, income, etc.) to classify agents into classes (e.g., creditworthy or not). Sometimes the attribute values are altered and/or hidden by agents to obtain a favorable but undeserved classification. Our main goal is to develop methods to thwart agents from hiding or distorting attribute values to obtain a favorable but incorrect classification. Intentionally altered attributes to obtain strategic goals have been studied. In this paper we develop methods that handle strategic hiding (i.e., nondisclosure) and then merge them with methods to thwart strategic distortion in the context of classification. [ABSTRACT FROM AUTHOR]

We study the problem where a decision maker needs to discover a classification rule to classify intelligent, self-interested agents. Agents may engage in strategic behavior to alter their characteristics for a favorable classification. We show how the decision maker can induce a classification rule that anticipates such behavior while still satisfying an important risk minimization principle. [ABSTRACT FROM AUTHOR]

Statistical learning theory provides a formal criterion for learning a concept from examples. This theory addresses directly the trade-off in empirical fit and generalization. In practice, this leads to the structural risk-minimization principle where one minimizes a bound on the overall risk functional. For learning linear discriminant functions, this bound is impacted by the minimum of two terms-the dimension and the inverse of the margin. A popular and powerful learning mechanism, support vector machines, focuses on maximizing the margin. We compare this to methods that focus on minimizing the dimensionality, which, coincidentally, fulfills another useful criterion-the minimum description length principle. [ABSTRACT FROM AUTHOR]

A procedure is developed for determining two-group linear discriminant classifiers that misclassify the fewest number of observations in the training sample. An experimental study confirms the value of this approach. [ABSTRACT FROM AUTHOR]

Many linear programming models have been proposed for performing discriminant analysis. Partial characterizations for unacceptable solutions have been presented and new models proposed to circumvent these problems. In this paper those conditions leading to unacceptable solutions for all two-group models are characterized. [ABSTRACT FROM AUTHOR]

Focuses on problems that may appear in mathematical programming models for linear discriminant analysis. History of the development of the discriminant analysis method; Problems involved in the analysis method's choice of objective; Existence of unbounded solutions in discriminant analysis.

Presents mathematical programming models for linear discriminant analysis. Applications of the models; Pathologies inherent in several of the models; Functions of discriminant analysis.