Table 6 Ablation study showing the individual and combined performance of CNN-based classifiers.
From: Automated multi-model framework for malaria detection using deep learning and feature fusion
Model | Accuracy | F1-Score | MCC | Key Observation |
|---|---|---|---|---|
CNN backbones only ⢠ResNet-50 ⢠VGG-16 ⢠DenseNet-201 | 95.77â96.32% | 95.75â96.30% | 91.96â93.15% | Establishes a strong deep-learning baseline. DenseNet-201 gives the best recall (SEN =â96.32%). |
Hybrid classical models only ⢠SVM ⢠LSTM | 96.40% ââ96.11% | 96.38% ââ96.09% | 93.09% ââ92.60% | Shows that the PCA-based feature-fusion vector, when fed to SVM or LSTM, already surpasses or matches the best single CNN. |
Ensemble layer ⢠Majority Voting (CNN +âSVM +âLSTM) | 96.47% | 96.45% | 93.35% | Delivers the global optimum across every metric |
