Anova-informed decision trees for voice applications over manets

Abstract

Both real-time multimedia and mobile networks present challenges ripe for new analysis techniques. We examine the applicability of statistical design of experiments and inductive learning theory in the prediction of delay for real-time audio transmissions over mobile ad hoc networks. Utilizing analysis of variance methods and simple decision tree agents, we find both significant factor interaction between traffic load and node mobility as well as a dramatic reduction in error percentage in prediction of end-to-end delay. © 2005 by International Federation for Information Processing.