Table 3 Ablation study: performance impact of key components in PersonalCareNet.
From: Personalized health monitoring using explainable AI: bridging trust in predictive healthcare
Model variant | Accuracy (%) | Precision (%) | Recall (%) | F1-Score (%) | AUC (%) |
|---|---|---|---|---|---|
PersonalCareNet (Full Model) | 97.86 | 96.2 | 95.4 | 95.8 | 98.3 |
w/o Feature Selection (No PCA) | 95.34 | 92.8 | 91.5 | 92.1 | 96.1 |
w/o Dropout Regularization | 94.76 | 91.7 | 90.2 | 90.9 | 95.2 |
w/o Attention Layer | 93.25 | 89.1 | 88.6 | 88.8 | 93.9 |
Base DNN (no enhancements) | 90.30 | 88.0 | 86.7 | 87.3 | 92.6 |
