Algorithm 1. Meta-Learner_Optimization

Input: Ensemble of deep learning classifiers {C1, C2,…, C7}, Training data D Output: Optimized weights for each classifier in the ensemble

1: Initialize weights Wi for each classifier Ci in the ensemble, i = 1 to 7, such that sum (Wi) = 1 2: For each training epoch or until performance converges do 3: For each classifier Ci in the ensemble do 4: Extract meta-features: Accuracy Ai, Loss Li, Confidence Level ConfLi from Ci using D 5:  Calculate performance score PSi for Ci using Ai, Li, ConfLi 6:  Update weight Wi for Ci based on PSi 7:  End For 8:  Evaluate ensemble performance on validation set using updated weights 9:  If ensemble performance has converged or improved minimally then 10:    Break from the loop 11:  End If 12: End For Procedure Calculate_Performance_Score(Accuracy Ai, Loss Li, Confidence Level ConfLi) 1:  Define a performance function F that considers Ai, Li, ConfLi 2:  Return performance score PSi = F(Ai, Li, ConfLi) Procedure Update_Weight(Performance Score PSi) 1:  Define a weighting strategy that adjusts Wi based on PSi 2:  Update Wi according to the defined strategy 3:  Normalize all weights Wi so that sum(Wi) = 1 Procedure Evaluate_Ensemble(Validation Data V) 1:  For each data point in V do 2:  Aggregate predictions from all classifiers using their weights Wi 3:  End For 4:  Calculate and return the overall performance of the ensemble on V