Prognostic Models Based on Linear Separability

摘要

Prognostic models are often designed on the basis of learning sets in accordance with multivariate regression methods. Recently, the interval regression and the ranked regression methods have been developed. Both these methods are useful in modeling censored data used in survival analysis. Designing the interval regression models as well as the ranked regression models can be treated similarly as the problem of linear classifier designing and linked to the concept of linear separability used in pattern recognition. The term linear separability refers to the examination of separation of two sets by a hyperplane in a given feature space.