Yazar "Akkoyun, Serkan" seçeneğine göre listele
Listeleniyor 1 - 7 / 7
Sayfa Başına Sonuç
Sıralama seçenekleri
Öğe Accurate prediction of gamow-teller beta-decay matrix elements via machine learning: Implications for nuclear structure(ELSEVIER, 2025) Yeşilkanat, Cafer Mert; Akkoyun, SerkanAccurate prediction of Gamow-Teller (GT) beta decay matrix elements [M(GT)] is essential for elucidating complex nuclear structure phenomena and understanding astrophysical processes. In this study, we employed five advanced machine learning models (Cubist, Support Vector Regression, Extreme Gradient Boosting, Random Forest, and Bayesian Regularized Neural Networks) to predict GT beta decay matrix elements in sd-shell nuclei, using experimental data from NNDC/ENSDF, NUBASE2016, and AME2016. This study systematically compared the predictive performance of traditional theoretical approaches (including the USDB, IM-SRG, CCEI, and CEFT) to that of advanced machine learning models trained based on experimental observations. Our primary objective was to determine whether data-driven models could achieve higher predictive accuracy than computationally expensive theoretical models by learning the complex and nonlinear relationships among experimental parameters that reflect nuclear structure and decay dynamics. The results demonstrate that the Cubist model achieves a significantly lower RMSE (0.073 in the full parameter modeling approach and 0.112 in the reduced parameter modeling approach) and high coefficients of determination (R2 = 0.901 and 0.919, respectively), thereby outperforming traditional methods. Furthermore, SHapley Additive exPlanations (SHAP) analysis revealed that a minimal set of critical nuclear parameters predominantly governs GT decay dynamics, thereby enhancing model interpretability without compromising predictive accuracy. Complementing these findings, an online calculator was developed to facilitate rapid, highfidelity predictions of GT matrix elements. Overall, our study demonstrates that a data-driven approach outperforms established theoretical models. More importantly, by identifying the minimal set of physical observables that govern GT transitions, our work provides crucial insights into the underlying physics of nuclear structure and offers a new benchmark for refining future theoretical models and astrophysical calculations.Öğe Applications of different machine learning methods on nuclear charge radius estimations(Institute of Physics, 2023) Bayram, Tuncay; Yeşilkanat, Cafer Mert; Akkoyun, SerkanTheoretical models come into play when the radius of nuclear charge, one of the most fundamental properties of atomic nuclei, cannot be measured using different experimental techniques. As an alternative to these models, machine learning (ML) can be considered as a different approach. In this study, ML techniques were performed using the experimental charge radius of 933 atomic nuclei (A ≥ 40 and Z ≥ 20) available in the literature. In the calculations in which eight different approaches were discussed, the obtained outcomes were compared with the experimental data, and the success of each ML approach in estimating the charge radius was revealed. As a result of the study, it was seen that the Cubist model approach was more successful than the others. It has also been observed that ML methods do not miss the different behavior in the magic numbers region.Öğe Estimation of fission barrier heights for even–even superheavy nuclei using machine learning approaches(Iop Science, 2023) Yeşilkanat, Cafer Mert; Akkoyun, SerkanWith the fission barrier height information, the survival probabilities of super-heavy nuclei can also be reached. Therefore, it is important to have accurate knowledge of fission barriers, for example, the discovery of super-heavy nuclei in the stability island in the super-heavy nuclei region. In this study, five machine learning techniques, Cubist model, Random Forest, support vector regression, extreme gradient boosting and artificial neural network were used to accurately predict the fission barriers of 330 even–even super-heavy nuclei in the region 140 ? N ? 216 with proton numbers between 92 and 120. The obtained results were compared both among themselves and with other theoretical model calculation estimates and experimental results. According to the results obtained, it was concluded that the Cubist model, support vector regression and extreme gradient boosting methods generally gave better results and could be a better tool for estimating fission barrier heights.Öğe Generation of fusion and fusion-evaporation reaction cross-sections by two-step machine learning methods(Elsevier B.V., 2024) Akkoyun, Serkan; Yeşilkanat, Cafer Mert; Bayram, TuncayIn order to obtain cross-sections of heavy-ion fusion and fusion-evaporation reactions, artificial neural networks, cubist, random forest, support vector regression, extreme gradient boosting, and multiple linear regression machine learning approaches were used separately in this study. The outcomes from these different methods that are obtained from the training carried out with the existing experimental data in the literature were compared. Furthermore, it has been observed that a two-step process yielded better results for determining the heavy-ion reaction cross-sections, after first estimating which approach would be better for which reaction. In this manner, the method for which the cross-section needs to be calculated is determined by the machine learning classification application, and predictions can be made using the machine learning regression application with the determined method. It has been concluded that the obtained results are in harmony with the experimental data and that the methods can be used safely. The obtained results are published on a web page that allows for online calculation of heavy-ion fusion and fusion-evaporation reaction cross-sections.Öğe Machine learning predictions for cross-sections of ⁴³,⁴⁴Sc radioisotope production by alpha-induced reactions on Ca target(Elsevier B.V., 2024) Akkoyun, Serkan; Yeşilkanat, Cafer Mert; Bayram, Tuncay??,??Sc radioisotopes are an alternative to ¹?F in positron emission tomography. ??,??Sc radioisotopes, which can be generated at low costs by irradiating inexpensive natural Ca with alpha particles, can be produced, and distributed in a central cyclotron facility due to their relatively long half-lives. Since there is limited experimental data on the cross-sections in the literature, in this study, cross-section predictions of the production of ??,??Sc radioisotopes with alpha particles on Ca target were carried out with different machine learning approaches. In order to improve the results, the feature engineering method was applied to the variables of the cross-section predictions. Moreover, predictions have been improved with Stacked Ensemble Learning (SEL) approaches, a complex methodology that leverages the predictive capabilities of multiple underlying models to build a higher-level metamodel. We found that the best results were obtained with Bayesian Regularized Neural Network, Support Vector Regression and Stacked Ensemble Learning methods.Öğe Neutron-alpha reaction cross section determination by machine learning approaches(Springer, 2024) Amrani, Naima; Yeşilkanat, Cafer Mert; Akkoyun, SerkanThis study focuses on leveraging powerful machine learning approaches to determine neutron- alpha reaction cross-sections within the 14–15 MeV energy range. The investigation utilizes an experimental dataset comprising measurements of 133 nuclei concerning (n, ?) reaction cross- sections. These data are divided into training and validation subsets, following established protocols, with 80% allocated for model training and 20% for testing. Key nucleus characteristics, including neutron number (N), mass number (A), and symmetry representation [(N-Z)²/A], were used as input variables for the machine learning models. SVR and XGBoost methods showed superior performance among the other machine learning methods used in the present study. In addition, a machine learning based online calculation tool was developed to estimate the reaction cross section.Öğe SMOTE-based data augmentation for accurate classification of neutron halo nuclei: A machine learning approach in nuclear physics(Elsevier B.V., 2025) Yeşilkanat, Cafer Mert; Akkoyun, SerkanNeutron halo nuclei exhibit unique structural features—such as extended matter distributions, large interaction cross sections, and unusually low separation energies—that offer valuable insights into the nature of nuclear forces, stability limits, and astrophysical nucleosynthesis. Traditional analytical methods face challenges in accurately characterizing these exotic systems, highlighting the need for advanced computational techniques. In this study, we propose a machine learning–based framework that leverages Synthetic Minority Over-sampling Technique (SMOTE)-based data augmentation to address class imbalance in neutron halo classification tasks. A comprehensive evaluation is conducted on eight widely used algorithms—AdaBoost, XGBoost, C5.0, Generalized Linear Models (GLM), k-Nearest Neighbors (kNN), Naive Bayes, Random Forest, and Support Vector Machines (SVM)—, assessing their predictive performance and computational efficiency. The results demonstrate that AdaBoost and XGBoost provide superior accuracy and stability, offering a robust approach to identifying potential neutron halo candidates. Additionally, we develop an interactive Shiny application for real-time classification, thereby strengthening the connection between data-driven methodologies and nuclear structure research. Overall, this work underscores the importance of data augmentation in nuclear physics and highlights the potential of machine learning-driven strategies for the automated identification of exotic halo nuclei, paving the way for more in-depth exploration of nuclear stability and structure.
