Supervised Machine Learning for Refractive Index Structure Parameter Modeling

The Hellenic Naval Academy (HNA) reports the latest results from a medium-range, near-maritime, free-space laser-communications-testing facility, between lighthouse of Psitalia Island and academy’s laboratory building. FSO link is established within premises Piraeus port, with path length 2958 m an average altitude 35 m, mainly above water. Recently, facility was upgraded through addition BLS450 scintillometer, which co-located MRV TS5000/155 system WS-2000 weather station. This paper presents preliminary optical turbulence measurements, collected 24 to 31 May 2022, alongside macroscopic meteorological parameters. Four machine-learning algorithms (random forest (RF), gradient boosting regressor (GBR), single layer (ANN), deep neural network (DNN)) were utilized for refractive-index-structural-parameter regression modeling. Additionally, another DNN used classify strength level turbulence, as either strong or weak. showed very good prediction accuracy all models. Specifically, ANN algorithm resulted in R-squared 0.896 mean square error (MSE) 0.0834; RF also gave highly acceptable 0.865 root (RMSE) 0.241. Gradient Boosting Regressor (GBR) 0.851 RMSE 0.252 and, finally, 0.79 0.088. DNN-turbulence-strength-classification model exhibited classification performance, given variability our target value (Cn2), since we observed predictive 87% model.