C. Gandía Tortosa, M. D. Molina Vila, M. J. Nueda Roldán
In the last decade classification methods have become in one of the main topics in the modern data analysis. The high number of variables available and specific types of distributional assumptions are challenging topics that scientists try to address with approaches as bagging, boosting or developing new statistical methods.
We address the classification problem formulating a linear problem, that looks for a hyperplane, H, which separates two groups. If such hyperplane does not exist, H will be found by minimizing the sum of all the possible infeasibilities. It results in a convex optimization problem for which we find an equivalent problem of linear programming that is solvable by applying the Karush, Kuhn and Tucker conditions.
We apply our method to RNA-seq data and compare results with SVM, support vector machine, that is a popular classification method also based on separating hyperplanes.
Palabras clave / Keywords: classification, linear programming, RNA-seq
Programado
Sesión J04 Métodos y Aplicaciones de la Investigación Operativa
31 de mayo de 2018 10:20
Sala 3
