Revista Egitania Sciencia - Volume 9 | ARTICLE

Title: EVALUATE THE PERFORMANCE OF AN ARTIFICIAL NEURONAL NETWORK WITH SPSS SOFTWARE

Author: Maria Cristina Canavarro Teixeira (ccanavarro@ipcb.pt), Nuria Ceular Villamandos (td1cevin@uco.es) e J. Maria Caridad y Ocerin (ccjm@uco.es)

Publication: Revista Egitania Sciencia - Volume 9

Abstract:
Neural Networks is the module of Statistical Package for Social Sciences (SPSS), which allows the users to make the appropriate modeling selections to their problems, in the estimation of artificial neural networks. These models recognize nonlinear patterns on data. To evaluate the performance of these models, the Model Summary of the Neural Networks module offers two error measures: the Sum of Squares Error and Relative Error. The main goal of this study is to evaluate the ability of the performance analysis of neural networks with SPSS. Various error measures that have been used over time to measure the predictive ability of neural networks are presented. We consider that the measures provided in this software are not enough, especially if we compare the neuronal models with other models, such as the multiple regression ones. We propose a way to overcome what we consider a limitation of SPSS, and also make some suggestions to include more information in that module.

Keywords: Neural Networks; SPSS; performance measures




               INSTITUTO POLITÉCNICO DA GUARDA | Av. Dr. Francisco Sá Carneiro, 50 | 6300 - 559 Guarda | Tel.+351271220100 | Fax +351271222690