We investigate the accurate description of $\alpha$-decay half-lives across a wide range of nuclei by integrating physical modeling with machine learning techniques. A double-folding potential model (DPM), constructed using a Skyrme-type effective nucleon-nucleon interaction that neglects momentum and spin dependence, is employed to calculate the $\alpha$-decay half-lives of 498 nuclei with $Z=61–118$. The model yields a root-mean-square (RMS) deviation of 0.58 from experimental data. To improve predictive accuracy, a support vector regression (SVR) method is applied. The SVR is optimized using the artificial bee colony algorithm and trained with the same nuclear input features {$Z, A, Q_{\alpha }, l$}. This approach reduces the RMS deviation to 0.44. Furthermore, by incorporating the SVR-predicted preformation factor into the DPM, the SVR-optimized DPM achieves a further reduction in RMS deviation to 0.41. The predictive performance of all three models is evaluated in the superheavy nuclear region, and their results are used to estimate the $\alpha$-decay half-lives of isotopes with $Z$= 117 and 118. These results offer reliable theoretical references for future experimental exploration of superheavy elements.

Close

PrevNext

View figure in article

• Alpha decay
• Lifetimes & widths
• Nuclear structure & decays
• A ≥ 220