Finding Top- Covering Irreducible Contrast Sequence Rules for Disease Diagnosis

Algorithm 1

The MineTopIRs Algorithm.

Input: a gene expression dataset ; the required number of rules ; the equivalent threshold ; the support threshold ;

the confidence threshold

Output: All top- covering irreducible contrast sequence rules for each sample with class label

()  Convert dataset into the EWave model , w.r.t. ;

()  Construct Head-Tail matrix;

()  Initiate a list of rules with both support and confidence values of 0, for each sample with class

label ;

()  Initiate the rule candidate set Candi_R with all 1-size sequence rules;

()  Call breathfirst_search(candi_R, , , 1);

()  Return for every with class label ;

Function: breathfirst_search(candi_R, , , )

()  while candi_R  do

()   foreach -size rule    generated based on the  -size rules in candi_R do

()    if ∀ -size subrule of    exists in candi_R then

()     if supp() > then Pruning rule 1;

()      if conf() < then

()       Pruning rule 2;

()       add into candi_R;

()      else

()       Check the th covering rule for each sample to find

       the lowest confidence minconf and the corresponding support sup;

()      if (conf() > minconf)∨(conf() = minconf∧supp() ≥ sup) then

()      Pruning rule 3;

()      Update for each sample with

        based on Definitions 18 and 20;

()   end

()  end

()  Delete all the -size rules in candi_R;

()  ++;

() end