Prediction of long-term hospitalisation and all-cause mortality in patients with chronic heart failure on Dutch claims data: a machine learning approach

Table 1 Hyperparameters used in the several models

Backward logistic regression Select criterion: significance level Stay significance level: 0.05	Elastic Net Alpha: 0–1 stepped by 0.1 Lambda*: 0.001 to 100.000 in 80 exponential increasing steps Folds for cross validation: 10
Random forest Max trees Max depth Split criterion: Gini	Neural network Type: fully connected feed forward Architecture: 1–3 hidden layer with 10–100 nodes

^*The lambda sequence is created using the following formula 10 ^ seq (from = − 2, to = 5, by = 0.1). This generates 71 values from 0 to 100.000

ISSN: 1472-6947