Use of VNS and TS in classification: variable selection and determination of the linear discrimination function coefficients

In this paper, we discuss the problem of variable selection and the determination of the coefficients for these variables that provide the best linear discrimination function with the objective of obtaining a high classification success rate. Given the relation that exists between both problems, they will be performed simultaneously. In order to resolve them, two algorithms based on the metaheuristic approaches variable neighbourhood search and tabu search have been designed. These methods have proved to obtain significantly better results than two of the most classic methods for obtaining discriminant linear functions: classic discriminant analysis and logistic regression.