Table 4 Performance in the test split of the RF model in the five-class study. Results in more detail are found in Supplementary Table S10. Results from 5-fold cross-validation are given as the balanced accuracy of the model in mean (%) ± standard deviation (%) format.
From: Machine learning-based approaches for cancer prediction using microbiome data
Balanced accuracy | |
|---|---|
Experiment 1: RF | 60.37 ± 2.93 |
Experiment 2: RF + Dimensionality Reduction | 61.25 ± 4.39 |
Experiment 3: RF + Feature engineering | 64.01 ± 2.25 |
Experiment 4: RF + Dimensionality reduction + Oversampling | 60.38 ± 3.12 |
Experiment 5: RF + Feature engineering + Oversampling | 67.31 ± 3.93 |
