Integrating Correlation-Based Feature Selection and Clustering for Improved Cardiovascular Disease Diagnosis

Algorithm 1

Proposed feature selection algorithm using reversed correlations

Input: F = f₁, f₂, f₃, … f_n/ set of all the features /;

   P/ statistical significance level /;

   R/ a threshold for correlation coefficient levels ^∗/;

   N/ the maximum of features for the subset/;

Output: F_s/ selected subset of features /;

   (1) Initialize F_s with feature f_j ϵ F that is the least correlated with other ones;

   (2) do

   (3) Compute C_ij(F_s, F \ F_s) as a vector of correlation coefficients between F_s and each f_i ϵ {F \ F_s};

   (4) Choose f_j ϵ {F \ F_s} with the lowest value of correlation coefficient in a vector C_ij(F_s, F \ F_s);

   (5) Include f_j in F_s

   (6) while (s < N AND p > P AND C_ij(F_s, F \ F_s) < R).