Geometric Dilution of Precision (GDOP) is often used for selecting optimum GPS satellites to meet the desired positioning precision. The conventional matrix inversion method for GPS GDOP calculation has a large amount of operation, which would be a burden for real-time application. Previous studies have partially solved this problem with Neural Network (NN). Though NN is a powerful function approximation technique, it needs time and power costs for training. Also, the trained NN may not be applicable to other data that are deviated from the training data. This paper employs Genetic Algorithm (GA) approach for GPS GDOP approximation and classification, without complicated matrix inversion. Using the proposed method, the calculations costs of GPS GDOP can be reduced. The computer simulations show that the proposed method with much reduced computational complexity has good performance.