Portal de Periódicos da CAPES

A New Partial Least Square Method Based on Elman Neural Network

2014; Volume: 4; Issue: 4 Linguagem: Inglês

2165-8994

Elif Bulut, Erol Eğrioğlu,

Neural Networks and Applications

Partial least square regression (PLSR) is a latent variable based multivariate statistical method that is a combination of partial least square (PLS) and multiple linear regressions. It accounts for small sample size, large number of predictor variables, correlated variables and several response variables. It is almost used commonly in all area. But in some complicated data sets linear PLS methods do not give satisfactory results so nonlinear PLS approaches were examined in literature. Feed forward artificial neural networks based nonlinear PLS method was proposed in the literature. In this study, the method of nonlinear PLS is improved to make a suggestion of a new nonlinear PLS method which is based on Elman feedback artificial neural networks. The proposed method is applied to data set of “30 young football players enrolled in the league of Football Players who are Candidates of Professional Leagues” and compare with some PLS methods.