Predicting BER value in OFDM-FSO systems using Machine Learning techniques
Ranim Younes, Fadi Ghosna1, Mohammad Nassr, Mohammad Anbar, Hamid Ali Abed Alasadi
Descargar artículo
Información básica
Volumen
V55 - N4 / 2022 Ordinario
Referencia
51114
DOI
http://dx.doi.org/10.7149/OPA.55.4.51114
Idioma
English
Etiquetas
Free Space Optics (FSO), OFDM, Optisystem, Machine Learning Algorithms (MLA), Prediction, Regression, Support Vector Regression (SVR), Decision Tree (DT), Random Forest (RF), Bit Error Rate (BER).
Resumen
Recently, Free Space Optics (FSO) has emerged as new technology for transmission through atmosphere. It is difficult to implement FSO systems under bad weather conditions such as fog and rain and so on. These conditions cause deterioration in the FSO system signal. Thus OFDM technology has been used to enhance system performance and to overcome signal weakness due to weather conditions. Machine Learning Algorithms (MLAs) are good prediction tools which can improveperformances of communication networksin general. In this work, three of machine learning algorithms (namely: Support Vector Regression (SVR), Decision Tree (DT), Random Forest (RF)) have been used to estimate the value of Bit Error Rate (BER). A data set has been obtained from Optisystem v.15 for training and testing MLAs modelsunder different weather conditions (Fog, Rain, Clear). The obtained results show that SVR algorithm cannot be used to predict BER value in the OFDM-FSO system. RF and DT algorithms gave approximate results. RF gave better accuracy where it has the greatest value of determination coefficient (R2) and the smallest value of Mean Square Error (MSE) compared to other algorithms.