\(\alpha \)-decay half-life predictions for superheavy elements through machine learning techniques

The stability and synthesis of superheavy nuclei are critically influenced by the accurate prediction of \(\alpha \)-decay half-lives. As an alternative to traditional models and empirical formulae, we employ the XGBoost machine learning algorithm for predicting the \(\alpha \)-decay half-lives of superheavy nuclei. For training the machine learning algorithm, the experimental half-lives of 344 nuclides in the mass range of 106 \(\le A \le 261\) and atomic numbers \(52 \le Z \le 107\) are used. Intricate correlations between nuclear features (Q value of the decay, mass, charge, neutron numbers) and half-lives are developed while training the XGBoost model with existing experimental data. The model performance is then assessed by comparing the predictions with experimental data and other empirical estimates. The trained model is found to have the least mean square deviation with respect to other empirical formulae. The trained model is then used to calculate the half lives of superheavy nuclei. The obtained results indicate that, in the superheavy element (SHE) region, XGBoost makes very effective predictions for the \(\alpha \)-decay half-lives. The impact of physics features is demonstrated with SHAP (SHapley Additive exPlanations) summary plots.