Incorporating Topic Assignment Constraint and Topic Correlation Limitation into Clinical Goal Discovering for Clinical Pathway Mining

Algorithm 1

Gibbs sampling of CDG-LDA.

     Input : A dataset of clinical days.

    Topic number .

    Hyper-parameters α and β.

    Iterations δ for updating relevant topic list.

     Output: The multinomial distribution ϕ and θ.

1 Calculate IDF for each clinical activity;

2 Initialize the topic number upper bounder

  for clinical activity with value v according to its IDF;

3 Initialize the relevant topic list for

  clinical activity with value v;

4 Sample for each group, and Initialize all the count

  parameters and accordingly;

5 for to K do

6      if then

7     for to do

8     if then

9        Update ;

10      for to do

11     for to do

12     for to do

13         ;

14         ;

15         Sample topic index from

        according to (6);

16         ;

17 Calculate and return ϕ (based on 8) and β;