An Evaluation of Neural Network Architecture Performance in Wireless Geo-Location

Abstract

Wireless geo-location applications require robust algorithms that are capable of locating and/or tracking wireless users requesting the service. To this effect, the performance of three neural network architectures has been evaluated through simulation to determine the optimal performance algorithm that can be applied to these new applications, such as location based-services (LBS). The results indicate that neural networks having self-organizing characteristics quickly learn to adapt to the rapid changing radio environment as opposed to other architectures which take much longer. Typical figures indicate that this family of neural networks reaches performance advantages of 45% and above when compared to other neural families making then the ideal candidates for such applications.