Performance of the proposed model with Gaussian filtering across five cross-validation folds (I-V). Metrics include Accuracy, Balanced Accuracy, Recall (TPR), Specificity (TNR), Precision (PPV), NPV, F1-Score, and Cohen’s Kappa, with averages in the final column. Compared to the baseline (Table 4), the model achieves higher average accuracy (77.3 % vs. 72.4 %) and balanced accuracy (82.5 % vs. 76.9 %), while maintaining very high specificity (97.4 %). This indicates overall improved classification performance and better handling of class imbalance.

Performance of the proposed model with Gaussian filtering across five cross-validation folds (I-V). Metrics include Accuracy, Balanced Accuracy, Recall (TPR), Specificity (TNR), Precision (PPV), NPV, F1-Score, and Cohen’s Kappa, with averages in the final column. Compared to the baseline (Table 4), the model achieves higher average accuracy (77.3 % vs. 72.4 %) and balanced accuracy (82.5 % vs. 76.9 %), while maintaining very high specificity (97.4 %). This indicates overall improved classification performance and better handling of class imbalance.