Combining Measurement Data and Erlang-B Formula for Blocking Prediction in GSM Networks
Pasi Lehtimäki, Kimmo Raivio · Scandinavian Conference on AI · 2008
In cellular network performance optimization, it is important to be able to predict the amount of user-experienced quality problems such as call blocking with alternative modifications to the network configuration. In this paper, a method based on mathematical optimization and knowledge representation to predict the amount of blocking in GSM networks is presented. The method is based on exploiting the available statistical measurement information describing the individual characteristics of the network elements. In addition, application domain knowledge about blocking is included to the model by using the well known Erlang-B formula to establish a mapping between blocking and existing network measurements. The results of the experiments show that the proposed method performs better than the comparison method based on basic Erlang-B formula and raw measurement data.