ESTIMATION OF COSMIC RADIATION DOSE RECEIVED ON FLIGHT BY MEANS OF DIFFERENT MACHINE LEARNING METHODS AND CARL-7A CALCULATIONS

Abstract

Sixteen machine learning methods including K-Nearest-Neighbor, Random Forest, Additive Regression, Linear Regression and M5P etc. were used to estimate the cosmic radiation dose for different international flights related with Istanbul and Ankara Airports in Turkey. Latitude, longitude and depth were used as inputs to the developed models, and the output variable is the dose rate. In order to evaluate accuracy of the developed models, five statistical indicators; correlation coefficient (R), mean absolute error (MAE), root mean squared error (RMSE), relative absolute error (RAE) and root relative squared error (RRSE) were compared. The results showed that K-Nearest-Neighbor (k-NN) model approach provides a high performance and lower error to predict the dose rate. Besides machine learning methods, the dose values in the flights were calculated with the CARI-7A software and the results obtained with both methods were compared and seen that they were in good agreement.