Table 5 Performance of the 5-year prediction model by deep neural network with autoencoder tree using repeated measured parameter.
Imputation method | Validation method | Validation ratio | Test set size | Main algorithm | Hyperparameters* | Test performance |
|---|---|---|---|---|---|---|
MICE/CART | One validation set | 0.1 | 174 | Decision Tree | cp = −1/maxdepth = 4 | 0.8012 |
MICE/CART | 174 | Logistic Regression | Nothing | 0.8045 | ||
MICE/CART | One validation set | 0.1 | 174 | Ridge | /lambda = 0.02 | 0.8236 |
MICE/CART | One validation set | 0.1 | 174 | Lasso | lambda = 0.02 | 0.8193 |
MICE/CART | One validation set | 0.1 | 174 | Bagging | nbagg = 30 | 0.8332 |
MICE/CART | One validation set | 0.1 | 174 | Random Forest | ntree = 500 | 0.8381 |
MICE/CART | One validation set | 0.1 | 174 | Neural Networks | hunits = [8] | 0.8066 |
Autoencoder | One validation set | 0.1 | 174 | Neural Networks | AE hunits = [16, 16]/FC hunits = [16] | 0.8419 |
Autoencoder | One validation set | 0.1 | 174 | Long Short-Term Memory networks | LSTM hunits = 16/AE hunits = [16, 16]/FC hunits = [16] | 0.8582 |
