Movement Prediction Using Bayesian Learning for Neural Networks

Abstract

A technique for reducing the wireless cost of tracking mobile users with uncertain parameters is developed in this paper. Such uncertainty arises naturally in wireless networks, since an efficient user tracking is based on a prediction of its future call and mobility parameters. The conventional approach based on dynamic tracking is not reliable in the sense that inaccurate prediction of the user mobility parameters may significantly reduce the tracking efficiency. Unfortunately, such uncertainty is unavoidable for mobile users, especially for a burst mobility patterns. In this paper, we present a novel hybrid Bayesian neural network model for predicting locations on Cellular Networks (can also be extended to other wireless networks such as WI-FI and WiMAX). We investigate different parallel implementation techniques on mobile devices of the proposed approach and compare it to many standard neural network techniques such as: Back-propagation, Elman, Resilient, Levenberg-Marqudat, and One-Step Secant models. Bayesian learning for Neural Networks predicts location better than standard neural network techniques since it uses well founded probability model to represent uncertainty about the relationships being learned. The result of Bayesian training is a posterior distribution over network weights. We use Markov chain Monte Carlo methods (MCMC) to sample N values from the posterior weights distribution.