Table 2.

Pseudocode of the ImputeEHR algorithm.

Algorithm: ImputeEHR algorithm
Require: X is n × m-dimensional data matrix, with stopping criterion γ
1. Make initial guess using mean or median imputation for missing values;
2. k ← A sorted indices vector according to t he amount of missing values of
column X;
w.r.t. increasing amount of missing values;
3. While not γ do
4.   $X_{old}^{imp}\leftarrow$ Store previously imputed matrix;
5.   for s in k do
6.      Fit a LightGBM or Xgboost : $y_{obs}^{(s)}~X_{obs}^{(s)}$;
7.      Predict $y_{miss}^{(s)}$ using $X_{miss}^{(s)}$;
8.      $X_{new}^{imp}\leftarrow$ update imputed matrix from $y_{miss}^{(s)}$;
9.   end for
10.   Update γ
11.   end while
12. Return Matrix X;