<![CDATA[Volcanic eruption disasters occur frequently in Indonesia due to the high density of active volcanoes, posing persistent risks to surrounding communities and infrastructure. Effective mitigation of these hazards is challenged by limitations in monitoring systems, particularly related to instrumentation coverage and the availability of expert human resources. One critical aspect of volcanic monitoring is the accurate classification of seismic activity, which reflects subsurface volcanic processes and supports timely hazard assessment. This study addresses the challenge of reliably classifying volcanic seismic events by proposing an integrated framework that combines autocorrelation-based signal characterization with Support Vector Machine (SVM)–based multi-class classification, supported by Z-score normalization during data preprocessing. The framework is designed to enhance feature consistency and robustness against noise commonly present in volcanic seismic signals. To evaluate its effectiveness, three SVM kernel functions—linear, polynomial, and radial basis function (RBF)—are systematically assessed under identical experimental conditions. The results demonstrate that the polynomial SVM kernel with a degree of two provides the most reliable classification performance, achieving an accuracy of 0.9605. In addition, the application of Z-score normalization substantially improves model stability and overall performance across all kernel configurations, indicating that feature scaling plays a critical role in SVM-based seismic classification. Performance variations among kernels suggest that non-linear feature representations are better suited to capture the complex characteristics of volcanic seismic signals, while classification errors are primarily influenced by class imbalance in underrepresented event types. These findings indicate that the proposed framework effectively supports automated volcanic seismic signal analysis and has the potential to enhance the reliability of seismic-based volcanic activity monitoring.]]>
