Table 3 Comparison of different machine learning classifers.
From: Enhancing noninvasive pancreatic cystic neoplasm diagnosis with multimodal machine learning
Models | Accuracy | Precision | Sensitivity | Specificity | ROC-AUC |
|---|---|---|---|---|---|
NaiveBayes | 0.6203 ± 0.0296 | 0.6394 ± 0.0237 | 0.6419 ± 0.0264 | 0.8741 ± 0.0116 | 0.7923 ± 0.0897 |
SVM | 0.5188 ± 0.0288 | 0.3670 ± 0.0346 | 0.3727 ± 0.0159 | 0.8049 ± 0.0058 | 0.6825 ± 0.0927 |
Decision Tree | 0.7878 ± 0.0087 | 0.7639 ± 0.0189 | 0.7648 ± 0.0199 | 0.9256 ± 0.0024 | 0.8755 ± 0.0522 |
CatBoost | 0.8964 ± 0.0260 | 0.9070 ± 0.0291 | 0.8698 ± 0.0206 | 0.9607 ± 0.0084 | 0.9272 ± 0.0259 |
XgBoost | 0.8911 ± 0.0246 | 0.8876 ± 0.0271 | 0.8681 ± 0.0270 | 0.9599 ± 0.0088 | 0.9243 ± 0.0155 |
Random Forest | 0.8680 ± 0.0101 | 0.8850 ± 0.0245 | 0.8221 ± 0.0184 | 0.9493 ± 0.0044 | 0.9065 ± 0.0432 |
Multilayer Perceptron | 0.7665 ± 0.0298 | 0.7474 ± 0.0315 | 0.7334 ± 0.3155 | 0.7335 ± 0.0370 | 0.8512 ± 0.0417 |
DNN | 0.8083 ± 0.0141 | 0.8007 ± 0.0208 | 0.7851 ± 0.0261 | 0.9326 ± 0.0054 | 0.901 ± 0.0153 |
CNN | 0.8135 ± 0.0621 | 0.8023 ± 0.0981 | 0.8471 ± 0.0339 | 0.8382 ± 0.0635 | 0.8789 ± 0.0313 |
Transformer | 0.8294 ± 0.0509 | 0.8374 ± 0.0387 | 0.8146 ± 0.0823 | 0.8070 ± 0.0143 | 0.9113 ± 0.0231 |
Multimodal Deep Forest | 0.9126 ± 0.0170 * | 0.9174 ± 0.0250* | 0.8891 ± 0.0202* | 0.9653 ± 0.0088* | 0.9584 ± 0.0214* |
Improvement | 1.8% ~ 75.9% | 3.35% ~ 149% | 2.21% ~ 138% | 0.47% ~ 31.6% | 3.36% ~ 40.4% |
