Data mining for indoor wave propagation model calibration

Accurate Radio wave propagation modeling has been for a longtime an important area of research and development due to their effect on network cost and deployment. In literature, many propagation models had been proposed and classified as empirical or deterministic. Empirical models are attractive since they are simple to use with low computational load. Their major drawback is the need of an adjustment to each new environment; this operation is called calibration or tuning. The outline of this paper is to define and compare two methods for indoor model tuning. The chosen model to tune is called Cheung model, which is an amelioration of multiwall one. The first used method for tuning is based on multi linear regressions theory, the second one is based on neural networks. To accomplish this task, measurements campaign has been performed in the higher school of communication of Tunisia (SupCom) building in 900, 1800, 2100 and 2400 MHz bands.