The determination of the mass composition of the ultra high energy cosmic rays is essential to many open questions in astroparticle physics. However the identification of the cosmic ray primary particle is a complex task due to several difficulties such as the large fluctuations in the shower development and the low number of experimental measurables. We present a proposal for composition studies applying multivariate analysis to make use of several parameters extracted from the longitudinal development of the showers to improve primary particle identification. Measurable features of the CR shower longitudinal profile such as the Nmax, Xmax, asymmetry, kurtosis, and skewness were combined using linear discriminant analysis (LDA). Studies were done using cosmic ray showers simulated by the CONEX code considering gamma, proton, helium, carbon and iron as primary particles.4HE PART 1531534Du, A., (2007) Astroparticle Physics, 27, p. 61Ui, A.R., (2005) Astrophys. Journal, 622, p. 910C. F., et al. astro-ph 0703582P. T., et al. astro-ph/0411260Su, O., (2006) Nuc. Phys. B (Proc. Sup), 151, p. 143De Souza, V., Phys. Rev. D, 73, p. 043001Au, J.R., Wu, W.D., Applied Multivariate Statistical Analysis, , Prentice HallJu, A., (2007) Astropart. Phys., 27, p. 155Mu, A., (2006) Nuclear Instruments and Methods A, 567, p. 107