3D Transition Matrix Solution for a Path Dependency Problem of Markov Chains-Based Prediction in Cellular Networks

Handover (HO) management is one of the critical challenges in current and future mobile communication systems due to new technologies being deployed at a network level, such as small and femtocells. Because of the smaller sizes of cells, users are expected to perform more frequent HOs, which can increase signaling costs and also decrease user's performance, if a HO is performed poorly. In order to address this issue, predictive HO techniques, such as Markov chains (MC), have been introduced in the literature due to their simplicity and generality. This technique, however, experiences a path dependency problem, specially when a user performs a HO to the same cell, also known as a re-visit. In this paper, the path dependency problem of this kind of predictors is tackled by introducing a new 3D transition matrix, which has an additional dimension representing the orders of HOs, instead of a conventional 2D one. Results show that the proposed algorithm outperforms the classical MC based predictors both in terms of accuracy and HO cost when re-visits are considered.