Mathematical programming in discriminant analysis for forecasting

For the first time the method of linear programming in the framework of discriminant analysis is applied to forecasting the life of products. The model of the separating surface is augmented by the factor of time $t$ which yields a kind of compositional model which reveals the dependence of the surface on time and on the vector $X$ of controlled variables. This model expands the potential of discriminant analysis from a «crude» forecast of the state to «pointwise» evaluation of the product longevity.