Table 1 The performance of the Langbehn formulas and the ML models to predict the AAO using participants with a narrow (n: 41–56) and wider (w: 36–59) CAG repeat size (caghigh)
Model | MAE | RMSE | R2 | |||
|---|---|---|---|---|---|---|
n | w | n | w | n | w | |
LGBM | 5.26 | 5.46 | 6.87 | 7.16 | 0.60 | 0.63 |
CatBoost | 5.29 | 5.49 | 6.91 | 7.20 | 0.60 | 0.62 |
Linear regression | 5.45 | 5.77 | 7.07 | 7.50 | 0.58 | 0.59 |
Langbehn refitted | 5.38 | 5.74 | 7.09 | 7.57 | 0.58 | 0.58 |
MLP | 5.46 | 5.77 | 7.14 | 7.55 | 0.57 | 0.58 |
XGBoost | 5.48 | 5.65 | 7.15 | 7.41 | 0.57 | 0.60 |
Langbehn | 5.57 | 5.91 | 7.23 | 7.88 | 0.56 | 0.55 |
Linear SVM | 5.68 | 5.96 | 7.35 | 7.75 | 0.55 | 0.56 |
Random forest | 5.85 | 6.03 | 7.63 | 7.89 | 0.51 | 0.55 |
Knn | 6.18 | 6.37 | 8.11 | 8.37 | 0.45 | 0.49 |