Home / Current Issue / Paper 1722944
Developing Kernel-Aware Explainability Techniques for Interpretable Support Vector Machines
Subject area: Science,Engineering and Technology · Area of research: Vector Machines
DOI: 10.64388/IREV10I3-1722944
Abstract
Artificial Intelligence (AI) systems are increasingly used for classification and decision support, but many machine learning models provide limited insight into how predictions are generated. This problem is particularly important in sensitive domains where users need to understand and trust automated decisions. Support Vector Machines (SVMs) provide strong classification performance, but nonlinear kernel functions make their decision processes difficult to interpret. This study developed four kernel-aware explainability techniques for providing insights into SVM decision-making: Support Vector Proximity Visualization (SVPV), Kernel Contribution Decomposition (KCD), Feature Influence Plot (FIP), and Decision Region Mapping (DRM). The Breast Cancer Wisconsin Diagnostic Dataset was used to evaluate three SVM models based on Linear, Polynomial, and Radial Basis Function (RBF) kernels. The models achieved testing accuracies of 95.2%, 96.0%, and 97.4%, respectively, while ROC-AUC scores were 0.96, 0.97, and 0.99. SVPV provided information about the relationship between test observations and influential support vectors, KCD examined kernel and support-vector contributions, FIP identified influential features, and DRM visualized classification regions and decision boundaries. Comparison with LIME and SHAP indicated that the proposed techniques provided high interpretability and stability with moderate computational requirements. User evaluation produced mean scores of 4.3, 4.1, 4.5, and 4.4 out of 5 for SVPV, KCD, FIP, and DRM, respectively. The findings demonstrate that model-specific, kernel-aware explanations can improve the transparency of SVM-based classification while retaining strong predictive performance.
Keywords
Explainable Artificial Intelligence, Support Vector Machines, Kernel Methods, Interpretability, Feature Influence, Decision Visualization.
References
[1] F. Doshi-Velez and B. Kim, “Towards a rigorous science of interpretable machine learning,” arXiv preprint arXiv:1702.08608, 2017.
[2] C. Cortes and V. Vapnik, “Support-vector networks,” Machine Learning, vol. 20, no. 3, pp. 273–297, 1995.
[3] N. H. Barakat and A. P. Bradley, “Rule extraction from support vector machines: A review,” Neurocomputing, vol. 74, no. 1–3, pp. 178–190, 2010.
[4] M. T. Ribeiro, S. Singh, and C. Guestrin, “‘Why should I trust you?’: Explaining the predictions of any classifier,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 1135–1144, 2016.
[5] S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems, vol. 30, pp. 4765–4774, 2017.
[6] R. Guidotti, A. Monreale, S. Ruggieri, F. Turini, F. Giannotti, and D. Pedreschi, “A survey of methods for explaining black box models,” ACM Computing Surveys, vol. 51, no. 5, pp. 1–42, 2018.
How to cite this paper
@article{1722944,
author = {Osunniyi Segun James, Yusuf Mariam Ayomide},
title = {Developing Kernel-Aware Explainability Techniques for Interpretable Support Vector Machines},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {10},
number = {3},
pages = {4002-4006},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1722944.pdf},
abstract = {Artificial Intelligence (AI) systems are increasingly used for classification and decision support, but many machine learning models provide limited insight into how predictions are generated. This problem is particularly important in sensitive domains where users need to understand and trust automated decisions. Support Vector Machines (SVMs) provide strong classification performance, but nonlinear kernel functions make their decision processes difficult to interpret. This study developed four kernel-aware explainability techniques for providing insights into SVM decision-making: Support Vector Proximity Visualization (SVPV), Kernel Contribution Decomposition (KCD), Feature Influence Plot (FIP), and Decision Region Mapping (DRM). The Breast Cancer Wisconsin Diagnostic Dataset was used to evaluate three SVM models based on Linear, Polynomial, and Radial Basis Function (RBF) kernels. The models achieved testing accuracies of 95.2%, 96.0%, and 97.4%, respectively, while ROC-AUC scores were 0.96, 0.97, and 0.99. SVPV provided information about the relationship between test observations and influential support vectors, KCD examined kernel and support-vector contributions, FIP identified influential features, and DRM visualized classification regions and decision boundaries. Comparison with LIME and SHAP indicated that the proposed techniques provided high interpretability and stability with moderate computational requirements. User evaluation produced mean scores of 4.3, 4.1, 4.5, and 4.4 out of 5 for SVPV, KCD, FIP, and DRM, respectively. The findings demonstrate that model-specific, kernel-aware explanations can improve the transparency of SVM-based classification while retaining strong predictive performance.},
keywords = {Explainable Artificial Intelligence, Support Vector Machines, Kernel Methods, Interpretability, Feature Influence, Decision Visualization.},
month = {September},
doi = {https://doi.org/10.64388/IREV10I3-1722944}
}