Resumen

CAMEJO CORONA, Julio; GONZALEZ, Hector  y  MORELL, Carlos. Multitarget Regression Problem. A review for Big Data. Rev cuba cienc informat [online]. 2019, vol.13, n.4, pp. 118-150. ISSN 2227-1899.

In many cases regression problems with more than one objective feature can be present. In these cases, you can model as many regressors as output variables exist, which underestimates the conditional dependence between the variable output pairs considering each independent problem. Recently it has been shown that considering this dependency produces better results since in many problems the output variables yield results that are related to each other. The high computational cost of these algorithms, and the enormous amount of information stored in millions of databases, has resulted in excessively large processing times in the generation of these models, which implies the need to manage these problems from Big Data concept. The objective of this article is to provide an overview of the current state of the main regression proposals with multiple outputs and their possibilities of being reformulated to Large-Scale problems. Besides, the followed methodology by the Multiple Linear Regression already implemented in the Apache Spark platform is addressed. Finally, the main optimization techniques that use these methods and their variants from Big Data are exposed.

Palabras clave : Multi-target regression; Regression; Apache Spark; Big Data; Large-Scale; Optimization.

        · resumen en Español     · texto en Español     · Español ( pdf )