Neural Network and Data Envelopment Analysis to solve binary Classification Problems

In this paper using additive model in Data Envelopment Analysis (DEA) technique, a methodfor solving separable linear and nonlinear classification problems presented. This method willresult an accurate classification for solving binary classification problems with linearlyseparable classes. One key feature of this method is that in regards of using additive DEAmodel there is no need for positive data and thus it has less computational complexity. Notedthat, for solving separable nonlinear binary classification, using Gaussian basic functions, anew RBFN artificial neural network will be presented. In corresponding section of supervisor,instead of using a linear function, a procedure as additive model in DEA has been utilizedwhich is more stable with higher speed in successive iterations. Furthermore, two examplesalso used to demonstrate the validity of the presented method.

Keywords:

Data Envelopment Analysis , Neural Network , Radial Basic Function , Binary classification