OUP accepted manuscript

The most abundant products of the interaction between radiation and matter are low-energy electrons, and the collisions between these electrons and biomolecules are the main initial source of radiation-based biological damage. To facilitate the rapid and accurate quantification of low-energy electrons (0.1-10 keV) in liquid water at different site diameters (1-2000 nm), this study obtained ${\overline{y}}_{\mathrm{F}}$ and ${\overline{y}}_{\mathrm{D}}$data for low-energy electrons under these conditions. This paper proposes a back-propagation (BP) neural network optimized by the mind evolutionary algorithm (MEA) to construct a prediction model and evaluate the corresponding prediction effect. The results show that the ${\overline{y}}_{\mathrm{F}}$ and ${\overline{y}}_{\mathrm{D}}$ values predicted by the MEA-BP neural network algorithm reach a training precision on the order of ${10}^{-8}$. The relative error range between the prediction results of the validated model and the Monte Carlo calculation results is 0.03-5.98% (the error range for single-energy electrons is 0.1-5.98%, and that for spectral distribution electrons is 0.03-4.4%).