α-decay half-life predictions with support vector machine.

Area of Science:

• Nuclear Physics
• Computational Physics
• Machine Learning

Background:

• Predicting nuclear decay half-lives is crucial for understanding nuclear structure and reactions.
• Traditional methods often require extensive experimental data or complex theoretical calculations.

Purpose of the Study:

• To apply support vector machines (SVM) with a radial basis function (RBF) kernel for predicting nuclear alpha-decay half-lives.
• To evaluate the impact of various physics-derived features on predictive accuracy.
• To identify key features influencing alpha-decay half-life predictions.

Main Methods:

• Utilized a dataset of 2232 nuclear data points.
• Employed SVM with an RBF kernel.
• Incorporated physics-derived features including nuclear structure characteristics, liquid drop model terms, decay energies, and quantum numbers.
• Applied Shapley additive explanations (SHAP) to interpret model predictions.

Main Results:

• Achieved root mean square errors of 0.819 (set1) and 0.352 (set2), comparable to other machine learning methods.
• Identified parent nuclei as the most significant feature for predicting alpha-decay half-lives.
• Demonstrated the effectiveness of the RBF kernel in SVM for this task.

Conclusions:

• SVM with an RBF kernel is a powerful tool for predicting nuclear alpha-decay half-lives.
• Physics-derived features, particularly those of parent nuclei, are highly predictive.
• This approach offers a promising avenue for predicting half-lives of unstudied nuclei, advancing nuclear structure research.