Abstract
Indoor radio propagation prediction modeling has been for a long time an important area of interest of research and development. The accuracy of developed model depends on the number of considered phenomena and their formulations. In general proposed formulation can be deduced from either experimental measurements, and then obtained model can be classified as empirical one. Or it can be deduced from physical lows of wave propagation, and hence, the developed model can be considered as exact or deterministic model. In this paper we are interesting to develop a new empirical indoor propagation model in 2100MHz band based on neural network theory. The new model has the advantage of combining many phenomena and parameters such as wall and floor attenuation and their number. To accomplish this work, a large number of Continuous Wave (CW) experimental measurements are collected and divided into two random set, the first one serves for model training while the second one serves for model validation. The new model shows high performances in term of accuracy.