The capability of the classical Linear Discriminant Analysis based on Generalized Singular Value Decomposition (LDA/GSVD) deteriorates when dealing with unlabeled datasets because LDA requires predefined inputs and targets. In addition, the LDA/GSVD algorithm suffers from high computation cost due to its complex mathematical calculations and iterations. To address these problems, this study introduces Self-Organizing Map (SOM) as a new method in labeling datasets, and the development of an Artificial Neural Network-based algorithm to overcome the computational cost of LDA/GSVD. The results show that using SOM and ANN are effective in solving the problems of the traditional LDA/GSVD algorithm.