Algorithm 4 Hybrid Multi-filter approaches and Correlation-Based Redundancy

Input: D$\left\{{\mathrm{f}}_1,{\mathrm{f}}_2,\dots ,{\mathrm{f}}_{\mathrm{m}}\right\}$A dataset with m features, number of filter h, number of union filtered gene n, number of genes subset ($S)P$, classifier C

Output: optimal feature subset $T$

for $i$ $\in$ $\left\{1\dots , h\right\} do$

for $j \in \left\{1, \dots ., m\right\} do$

employ $W_i$ filter to compute the statistical scores of each gene $G_i$

end for

select the top-ranking score in each list $G_i$ and get a new gene list $W_i{G}_j$, $j=1, \dots , m$

end for

produce a new ranking list $R$ by aggregating the output $k$ filter methods $W_i{G}_j$, $i=1, \dots , k$ using union the operator.

R/* the union of the list genes */

for each candidate feature $f_i$ in $G$, compute the interaction between feature-feature and feature-class, to discard redundant features based on Correlation-Based Redundancy using $SU$.

Initialize $S$ = $\varnothing$,

For$i=1ton,j=i+1$

$f_i$first feature, $f_j$second feature

$calculateS{U}_{i,c},S{U}_{i,j}andS{U}_{j,c}$

if$S{U}_{i,c}\ge S{U}_{j,c}\&S{U}_{i,j}\ge S{U}_{j,c}$then

remove $f_j$ $i.e.,S⟵S-\left\{f_j\right\}$.

else

insert $f_j$ into output selected features list $S⟵S+\left\{f_j\right\},$

end if

End of For

Return optimal feature subsets $S$