Algorithm 1:

Pseudocode for Q-prioritization phase.

Input: Ws

Output: Relevant prioritized queries.

1 Scan all labeled Ws determined in “Segregate-weight score phase” ;

2 Compute relevant threshold Rth for each PD subset: $R_{th}=\frac{Norma{l}_{max(PD)}}{Coun{t}_{col}\times Rat{e}_{max}}$ ;

3 foreach disorder set D do Add all rate points Ri grouped by column. ;

4 Find mean μi for each column. ;

5 Designate μi as Prel ;

6 Rank all Prel in ascending order priority: $P^{rel}=\sum _{i=1}^S\frac{R_i}{S}$ ;

7 if $P^{rel}\le R_{th}$ then

8 | corresponding attribute column is irrelevant. ;

9 end

10 if $P^{rel}\ge R_{th}$ then

11 | corresponding column is relevant. ;

12 end

13 Relevant prioritized queries are retained. ;