Table 2.
Performance comparison with previously published methods.
| Model | Classification | Inputs | Accuracy | AUC | Reference |
|---|---|---|---|---|---|
| Random Forest with Chi-Square feature selection | Binary (CN, MCI/AD) | Gene expression and clinical data (no CS) | 0.65 | 0.65 | This work |
| AdaBoost model with no feature selection | Binary (CN, MCI/AD) | SNPs and clinical data (no CS) | 0.67 | 0.63 | This work |
| SVM model with MI feature selection | Binary (CN, MCI/AD) | SNPs and gene and clinical (with CS) | 0.95 | 0.94 | This work |
| Deep neural network (DNN) | Binary (CN, AD) | Blood gene expression | NA | 0.656 | Lee and Lee22 |
| SVM | Binary (CN, AD) | Blood gene expression | NA | 0.620 | Lee and Lee22 |
| BSWiMS-LASSO-RPART ensemble | Binary (CN, AD) | SNPs | 0.677 | 0.719 | Oriol et al.25 |
| Deep learning models (DL) | Binary (CN, MCI/AD) | SNPs | 0.66 | NA | Venugopalan et al.36 |
Evaluation datasets were derived from ADNI by the respective authors.
BSWiMS: bootstrap stage-wise model selection; LASSO: least absolute shrinkage and selection operator; RPART: recursive partitioning and regression trees.