Deep data analysis methods applied to hard switching in hybrid FSO/RF systems

Abstract

This paper deals with the prediction of RSSI (Received Signal Strength Indicator) parameter which is required for hard switching methods in hybrid FSO/RF (Free Space Optics/Radio Frequency) systems. To ensure almost 100 % system availability, it is necessary to predict this parameter correctly. It is affected by several atmospheric influences such as fog, rain, airborne particle concentration, haze, etc. It is these atmospheric influences that have degrading effects on the RSSI parameter. Therefore, it is necessary to investigate and record these influences and correlate them to find out which have the greatest influence on the RSSI parameter. With proper prediction, we can determine well in advance which link will be communicating at that moment. The FSO line is used as the primary one for communication because of its data rate from one side to the other. Thus, the RF line serves only as a secondary (backup) line. For the prediction of RSSI parameter, machine learning methods of Decission Trees and Decission Trees along with AdaBoost Regressor were used.