Neural networks as non-parametric classification statistical tools. The relationships between neural networks and statistical methods have been recently analysized (see e.g. Cherkassky, F riedman and Wechler, 1994; Ripley, 1996). In general terms neural networks have shown and equal or greater capacity to classify than statistical tools. Morever they do not need to satisfy the parametric asumptions of the stistical techniques. Simulations about simulated and real data are shown: multi-layer perceptrons versus logistic regresion and discriminant analysis statistical models are compared in classifications tasks, manipulating the correlation patterns within input variables (predictors) and between the input variables with the output variable (criterion). Results show that neural networks classify better than statistical tools both in simulated data as in empirical data (Gonzalez-Roma et al, 1999).
[1]
G. David Garson,et al.
A Comparison of Neural Network and Expert Systems Algorithms with Common Multivariate Procedures for Analysis of Social Science Data
,
1991
.
[2]
Ian F. Croall,et al.
Industrial Applications of Neural Networks
,
1992,
Research Reports ESPRIT.
[3]
Huan Liu,et al.
Book review: Machine Learning, Neural and Statistical Classification Edited by D. Michie, D.J. Spiegelhalter and C.C. Taylor (Ellis Horwood Limited, 1994)
,
1996,
SGAR.
[4]
Bonifacio Martín del Brío,et al.
Redes neuronales y sistemas borrosos: introducción teórica y práctica
,
1997
.
[5]
Brian D. Ripley,et al.
Pattern Recognition and Neural Networks
,
1996
.