Assessment of statistical models to predict ambient gamma radiation levels using principal component analysis

Abstract EN

This study aims to produce a fitting model for predicting and assessing ambient gamma level using statistical regression models based on principal component analysis.

This prediction is useful because it provides early warning to the population with highly gamma levels and avoids exposure to them.

The study improves the predictive power of multiple linear regression models using principal components as inputs for predicting the gamma level of the next day or the missing record values.

The developed models are compared with multiple linear regression model.

The performance accuracy of the compared models is evaluated using Coefficient of Determination (R2), Index of Agreement (IA), Mean Square Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Normalized Mean absolute Error (NMAE), Normalized Mean Bias (NMB) and Fractional Absolute Error (FAE).

The results show that using the principal components and their rotation as input variables improved and gave better values than that of the multiple linear regression model.

This improvement is due to the complexity reduction and the elimination of data multi-collinearity.