Algorithm 1: Balancing Complex Signals

Begin

inputs:d –PPDS //ordered dataset

RF—the learning algorithm

$\overline{\mathrm{Q}}$ = {e1 = f − measure, e2 = kappa, e3 = accuracy}–the set of evaluation metrics;

output:Optimal model

1. RDS = Boruta(PPDS) 2. ODS = Order_dataset(RDS) 3.  For (λ = 0, λ ≤ 0.5, λ = λ + 0.05) A. ds = totds; clds = empty; olds = empty B. (for k = 1ncolumn(ODS) a. If(not(factor(ODS[k]) then b. classlabl = count(ODS[end]) c. Repeat i. u = Q3 + (1.5 + λ) * IQR(claslabl, attrbt) ii. l = Q1 − (1.5 + λ) * IQR(claslabl, attrbt) iii. cl = cleandataset(l, u, ODS, claslabl, k) iv. cs = complexsignals(l, u, ODS, claslabl, k) v. clds = combinedataset(clds, cl) vi. csds = combinedataset(csds, cs) vii. next(classlabl) d. Until(end of classlabl) e ds = clds f tot_cs = combinedataset (tot_cs, csds); g csds = empty; h end if; // end of step i. C. End For // step-b 4. Clean_Models_λ = BuildModel(clds, RF, ntrees) 5. Complex_Models_λ = BuildModel(tot_cs, RF, ntrees) 6. ResultsClean = ResultsClean + addPer (testModel(TestData, Clean_Models_λ,)) 7. End For // step-1 8. DisplayGraphs (Result) 9. Models_Results = Clean_Models_λ, ResultsClean 10. ${\mathrm{Opt}}_{{\mathrm{Perf}}^{\mathsf{\lambda }}}=\frac{{\mathrm{Accuracy}}^{\mathsf{\lambda }}\ast 100}{\left(perc\_c{\mathrm{s}}^{\mathsf{\lambda }}\right)}$ 11. $\mathrm{Retutn}\left({{\mathrm{Opt}}_{\mathrm{Perf}}}^{\mathsf{\lambda }}\right)$